# Nsightify URL: https://nsightify.com/ ## Operational Intelligence Should Be the Default. Not a Privilege. URL: https://nsightify.com/about/ Nsightify turns existing camera infrastructure into real-time operational intelligence for construction, warehousing and manufacturing operations. About [Home](/) \> About # Operational Intelligence Should Be the Default. Not a Privilege. Nsightify was built to give operations and security teams the real-time visibility they need — without requiring new hardware, IT projects, or giving up control of where your footage lives. Computer Vision Edge Intelligence Day-One Value Zero Disruption Real-World Ready Operations First Who we are ## A Team Built for Real-World Operations Nsightify is built by a team with deep expertise in machine learning, computer vision, and production software engineering. We’ve spent years working at the intersection of AI and real-world operational environments — training models, optimizing inference pipelines, and building software that runs reliably in places where failure isn’t an option. We’re based in Los Angeles, California. your camera, now intelligent. Our story ## Why We Built Nsightify The AI industry has an accessibility problem. The most capable computer vision tools are locked behind expensive cloud platforms, complex deployments, and months of integration work that most businesses can't absorb. A warehouse manager in Ohio shouldn't need a six-month implementation project just to know if workers are wearing hard hats. A construction site foreman in Texas shouldn't have to hire a data science team to get real-time visibility across his job site. We built Nsightify to close that gap. Work with the infrastructure businesses already own. Deliver operational insight from day one — not at the end of a procurement process. And let businesses choose where the intelligence runs — fully managed in Nsightify Cloud (SaaS-first, the default), or entirely on their own network with our On-Premises Zero Trust deployment when compliance and data sovereignty demand it. The result is a platform that gives any operations or security team the same caliber of AI-driven intelligence that was previously available only to enterprises with dedicated R&D budgets — without the hardware overhaul or the long rollout. What we stand for ## Our Principles Outcomes Privacy Resilience ### Outcomes Over Features Every capability we build connects to a real operational result: fewer injuries, fewer incidents, better compliance records, faster response. We don’t ship features for their own sake. ### Zero Disruption, by Design Businesses shouldn’t have to change how they operate to benefit from AI. Nsightify plugs into existing infrastructure, delivers value immediately, and creates no new operational dependencies. ### Privacy as Architecture — Zero Trust With our On-Premises (Zero Trust) deployment, video footage never leaves the customer's network. Only lightweight metadata — event descriptions, timestamps, zone identifiers — is ever transmitted. For organizations that need full data sovereignty, it's built into the architecture, not bolted on as an afterthought. ### Resilient When It Matters Internet outages happen. Nsightify’s on-premises deployment continues operating through connectivity loss — because safety and security monitoring can’t depend on a Wi-Fi connection. When connectivity returns, everything syncs automatically. We're Onboarding Pilot Partners Now If you're an operations or security leader who's been watching the AI space and waiting for something that actually works in your environment — we'd like to talk. [Join the Pilot Program](/pilot-program/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Talk to the Team URL: https://nsightify.com/contact/ Get in touch with the Nsightify team. Book a demo, apply to our pilot program, or ask about deployment, integrations, and pricing. We respond to every message. Contact [Home](/) \> Contact # Talk to the Team Whether you're ready to book a demo, want to join the pilot program, or have a specific question about your deployment — we respond to every message. [info@nsightify.com](mailto:info@nsightify.com?subject=Message%20from%20your%20site) Not sure what to say? Click to start: * "We have \[X\] cameras across \[Y\] sites. What would Nsightify look like for us?" * "We need on-premises deployment. Is our setup compatible?" * "We want to join the pilot. What’s involved?" Done! Thanks for your message. We'll get back as soon as possible. What can we help with?\* * Book a demo * Join the pilot program * Ask a question * Custom deployment inquiry * Press / media * Other What can we help with?\* Book a demoJoin the pilot programAsk a questionCustom deployment inquiryPress / mediaOther Submit [LinkedIn ](https://www.linkedin.com/company/nsightify-com/) [X ](https://x.com/NsightifyCom/) [Instagram ](https://www.instagram.com/nsightify/) [Facebook ](https://www.facebook.com/people/Nsightify/61587584023067/) ## Come prepared, we will too For deployment questions or pilot inquiries, include a brief description of your operation and camera setup — we'll come prepared. Location * Los Angeles, California Reach us * [info@nsightify.com](mailto:info@nsightify.com?subject=Message%20from%20your%20site) Let's talk about your operation! We're onboarding a limited number of pilot partners. Tell us about your sites and camera setup, and we'll show you what Nsightify looks like for you. [Say Hello](#contact-form) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Your Cameras See Everything. Your Business Sees Nothing. URL: https://nsightify.com/ Turn the cameras you already own into real-time operational intelligence — 28 safety and perimeter detections, including PPE. No camera replacement. AI-powered video analytics # Your Cameras See Everything. Your Business Sees Nothing. Nsightify turns your existing camera infrastructure into real-time AI video analytics — physical safety, perimeter security, and more — without replacing a single camera. [Book a Demo](/contact/) [See How It Works](#how-it-works) NSIGHTIFY LIVE FEED // CAM-04 REC ![Nsightify seeing and understanding a wide warehouse view.](/_astro/wide-warehouse-view-feed.BL5yQmEZ_1b8hAw.webp) ![Nsightify detecting workers, PPE and a perimeter zone on a live warehouse feed in real time](/_astro/ppe-violation-warehouse-annotated-feed.HJ1Lmf8d_Z1fRWOm.webp) ![Nsightify detecting an intruder crossing a restricted perimeter zone in real time](/_astro/intrusion-detection-violation-annotated-feed.Cj79rvuH_26eLSv.webp) Warehouse View Physical Safety Perimeter Security Physical Safety Perimeter Security No Camera Replacement Day-One Value The problem ## Recording Isn't Protecting Most businesses have security cameras installed. But cameras that only record don’t protect your operations — they document what went wrong after the fact. Adding AI shouldn’t mean rebuilding your stack or handing over control of your footage. Most platforms force one model on you. Nsightify lets you choose. Run it fully managed in our cloud, or entirely on your own servers where video never leaves your network. Same platform, same cameras, your rules. What Nsightify does ## One Platform. Many Operations. Choose the capabilities that matter most. Deploy individually or activate all of them. Every solution runs on the same cameras, the same platform, the same dashboard. Physical Safety ### Protect Your Workforce Before Incidents Happen 12 detections PPE detection for hard hats and hi-vis vests, plus vehicles in pedestrian walkways, forklift near-misses, blocked exits, and a worker down — in real time, across construction sites, warehouses, and manufacturing floors. [Explore PPE Detection & Physical Safety](/solutions/physical-safety/) Perimeter Security ### Know Exactly Who Enters — and Who Shouldn’t 16 detections Detect unauthorized entry, tripwire crossings, tailgating, loitering, fence-climbing, and someone tampering with a camera — instantly, without a human watching every feed 24/7. [Explore Perimeter Intrusion Detection](/solutions/perimeter-security/) Roadmap ### Animal Intrusion, Fire & Smoke, Licence Plates Nsightify is a platform, not a point tool. Animal intrusion for solar farms and substations, unattended baggage, fire and smoke, and licence-plate recognition are all in development — and pilot partners shape what ships first. [Join the Nsightify Pilot Program](/pilot-program/) How it works ## Up and Running in Hours. Not Months. 01 ### Connect Your Existing Cameras Connect Nsightify to your IP cameras via RTSP. Any camera you already have works. No camera replacement, no rewiring, no downtime. 02 ### Choose Your Deployment Nsightify Cloud for teams that want simplicity. On-Premises (Zero Trust) on your own servers for security-sensitive environments. Your infrastructure. Your rules. 03 ### Activate Your Use Cases Turn on physical safety, perimeter security, and more — individually or together. Configure zones and alert thresholds in minutes. 04 ### Operational Intelligence in Real Time Alerts, event logs, and compliance dashboards — delivered from day one. Not at the end of a six-month implementation. Deployment ## Your Infrastructure. Your Rules. Nsightify Cloud ### Simplicity at Scale Fully managed. Accessible from anywhere. Ideal for multi-site operations that need centralized visibility across locations without managing local infrastructure. Nsightify On-Premises (Zero Trust) ### Intelligence That Stays On Your Network Run Nsightify entirely on your own server(s). No video uploads. No cloud dependency. Full data sovereignty — built in, not bolted on. Both options. One platform. The cameras you already own. How it fits together ## From Camera to Alert You keep your cameras. The only extra piece is the Nsightify Edge Relay — and only when your cameras can't reach the internet directly. 01 · Capture Your existing cameras Any RTSP / IP / CCTV. No replacement. When needed 02 · Connect Nsightify Edge Relay A pre-configured mini-PC — only if your cameras can't reach the internet. Managed Nsightify Cloud Fully managed. We run the AI for you. or Self-hosted Your own server(s) Zero-Trust, on-premises. Video never leaves your network. 04 · Act Real-time alerts & dashboard Instant notifications and a live operations view. 05 · Record Logs & compliance Event logs, compliance records, events pushed to your systems. Which deployment? ## Nsightify Cloud vs. On-Premises (Zero Trust) Nsightify Cloud On-Premises (Zero Trust) Where the AI runs Nsightify Cloud Our cloud On-Premises (Zero Trust) Your own server(s) Where video goes Nsightify Cloud Streamed to cloud On-Premises (Zero Trust) Stays on your network Hardware you provide Nsightify Cloud Cameras (+ Edge Relay if needed) On-Premises (Zero Trust) Cameras + server(s) Works offline Nsightify Cloud No On-Premises (Zero Trust) Yes Best for Nsightify Cloud Fast start, multi-site On-Premises (Zero Trust) Data sovereignty, compliance Integrations ## Detections Don’t Stay in the Dashboard. Alerting Events & API Custom builds [Let's Talk About Integrations](/contact/) ### Alerts Reach People Where They Already Are A detection is only useful if someone sees it. Alerts leave the platform and land in the channels your team already checks — so nobody has to sit in front of another dashboard waiting for something to happen. ### Every Event Is Available to Your Own Systems Detections are published as structured events over an API and outbound webhooks. Your developers can pull them into the software that already runs your operation and build exactly the workflow you want — without waiting on us. ### If the Connection Doesn’t Exist, We Build It No two operations run the same stack, and we don’t pretend a fixed connector list covers yours. Tell us what the events need to reach, and we scope the integration with you as part of onboarding. Industries ## Built for the Way Real Operations Work ### Construction Monitor PPE compliance and hazard zones across active job sites — in real time, on every camera, without manual walkthroughs. ### Warehouses & Logistics Keep vehicles out of pedestrian walkways, catch forklift near-misses, keep egress routes clear, and monitor yards and loading docks — across every floor and every shift. ### Manufacturing Continuous safety monitoring on production floors. Detect hazard-zone breaches, PPE violations, and a worker down — without slowing operations. ### Security Operations Give security teams actionable intelligence instead of passive footage. Fewer false alarms. Faster response. Better coverage across every site. Why Nsightify ## No Camera Replacement. No Disruption. Intelligence from Day One. No rip-and-replace Works with the CCTV and IP cameras you already own. No camera procurement, no rewiring, no operational downtime. Immediate value No data science team required. No months-long implementation. Nsightify is configured and delivering alerts on day one. Flexible deployment Nsightify Cloud for simplicity. On-Premises (Zero Trust) on your own servers for data sovereignty. The platform is the same — the choice is yours. Multi-use-case Physical safety, perimeter security, and more on a single platform. Not three different tools from three different vendors. Real-world ready Designed for the floor, the yard, and the gate — not a demo environment. Handles variable lighting, crowded scenes, and real operational complexity. See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader with cameras already in place — let's talk. [Book a Demo](/contact/) [Learn About the Pilot Program](/pilot-program/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## AI Video Analytics: Turn Existing Cameras Into Intelligence URL: https://nsightify.com/nsights/ai-video-analytics-turn-cameras-into-intelligence/ Your cameras already see everything — they just don't tell you anything. Here's how AI video analytics turns existing feeds into real-time intelligence. [Home](/) [Nsights](/nsights/) AI Video Analytics: Turn Existing Cameras Into Intelligence June 19, 2026 4 min. read # AI Video Analytics: Turn Existing Cameras Into Intelligence AI & Security AI Video AnalyticsPhysical SafetyPerimeter SecurityComputer Vision ![Security cameras mounted on an industrial perimeter fence overlooking a logistics yard at dusk, with AI data visualization overlay](/_astro/security-cameras-industrial-site-ai.D30p1adF_10nofQ.webp) Walk onto any construction site, warehouse, or manufacturing floor and you’ll find cameras. Dozens of them. Sometimes hundreds. They run all day, record everything, and fill storage drives nobody ever opens. Then something goes wrong. A near-miss at the loading dock. Someone slips past the gate. A worker steps into a hazard zone without a hard hat. And what happens next? Someone pulls the footage — after the fact — to confirm what already happened. That’s the problem in one sentence. Your cameras are recording history. They aren’t telling you anything while it still matters. There’s a better use for that infrastructure, and it doesn’t require ripping any of it out. ## Footage Is Evidence. Intelligence Is Something Else Entirely. Recorded video is a record. You review it after an incident to understand what went wrong, settle a dispute, or hand something to an insurer. It’s useful — but it’s always backward-looking. By the time you’re watching the clip, the moment to act has passed. Intelligence is different. Intelligence is the system watching the feed in real time and telling you the instant something needs your attention — while you can still do something about it. The technology that makes this possible is computer vision: software that reads each frame of a video stream, recognizes people, objects, zones, and behaviors, and flags what matters. It’s the same broad capability behind a self-driving car reading the road, pointed instead at the cameras you already have running. ## What “AI Video Analytics” Actually Means Now For a long time, “watching the cameras” meant a person in a back room staring at a wall of monitors, hoping to catch something in the half-second it crossed one of forty screens. It doesn’t scale. Attention drifts. Most feeds are never watched live at all. Modern AI video analytics replaces that with software that watches every feed, continuously, and never blinks. It can recognize when: * A person enters a restricted or hazardous area * Someone on the floor isn’t wearing required protective equipment * A forklift comes dangerously close to a pedestrian * A vehicle or individual crosses a perimeter they shouldn’t * Two people pass a badge-secured door on one credential * An emergency exit or egress path is blocked * An area is more crowded than it should safely be * A camera has been obstructed, repositioned, or has quietly frozen When it sees one of these, it doesn’t file the clip away for later. It sends an alert — in real time — to the people who need it. That’s the shift: from a system you consult after an incident to a system that helps you prevent one. ## So Why Don’t More Businesses Have This Already? Because most of them assume it means starting over. The old model of AI video analytics came bundled with proprietary cameras, dedicated hardware boxes, new vendors to manage, and a multi-month IT project to glue it all together. For an operations or security leader who’s already stretched thin, that’s an easy “not this quarter.” So the cameras keep recording into the void, and the intelligence sitting inside those feeds goes unused. That model is outdated. You don’t need new cameras to get intelligence out of the ones you own. ## How Nsightify Turns Your Existing Cameras Into Intelligence Nsightify is a SaaS-first AI video analytics platform. It reads the feeds from the cameras you already have and turns them into real-time operational intelligence — [physical safety](/solutions/physical-safety/) and [perimeter security](/solutions/perimeter-security/) today, with more on the way. Getting connected is straightforward. Your existing IP cameras connect directly to the Nsightify platform. Where a direct connection isn’t practical, a lightweight Nsightify Edge Relay streams those same feeds up to the platform for you. Either way, the analysis runs in Nsightify — there’s nothing heavy to install on-site and no proprietary camera to buy. From there, the platform does the watching: * **[Physical safety](/solutions/physical-safety/)** — PPE compliance, hazard-zone enforcement, forklift–pedestrian proximity alerts, and blocked-exit detection, monitored continuously instead of during a twice-a-day walkthrough. * **[Perimeter security](/solutions/perimeter-security/)** — unauthorized entry, loitering, fence-climbing, and abandoned-object detection, with alerts the moment something needs attention rather than a clip you find the next morning. Integrations are designed to be effortless, so the platform fits into how your team already works instead of forcing a new workflow around it. For most businesses, that’s the whole picture: a subscription, your existing cameras, real-time alerts. For enterprises and regulated industries with stricter requirements, Nsightify also offers a Zero Trust deployment — keeping data inside your own private network or environment when compliance and data sovereignty demand it. ## Value on Day One — Not After a Six-Month Rollout The reason this matters is speed. There’s no long training period, no data science team to hire, and no months-long implementation before you see anything useful. You connect the cameras you already own, and the platform starts surfacing intelligence from the feeds that were already running. The cameras were always watching. Now they can finally tell you something — while you can still act on it. * * * _Ready to see what your existing cameras have been missing? [Book a demo](/contact/?interest=physical-safety) and we’ll show you real-time AI video analytics running on infrastructure you already own._ **Keep reading:** * [You Don’t Need New Cameras to Get AI Video Analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) * [Real-time physical safety monitoring on cameras you own](/nsights/physical-safety-monitoring-real-time-ai/) More on this from Nsightify: [AI video analytics on existing IP and CCTV cameras](/solutions/). [Next](/nsights/physical-safety-monitoring-real-time-ai/) [Real-Time Physical Safety Monitoring on Cameras You Own](/nsights/physical-safety-monitoring-real-time-ai/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## BIPA and Video Analytics: Where Biometric Collection Starts URL: https://nsightify.com/nsights/bipa-video-analytics-biometric-identifier/ BIPA video analytics risk sits in features, not cameras. What counts as a biometric identifier in Illinois, what 2026 rulings changed, and what to audit. [Home](/) [Nsights](/nsights/) BIPA and Video Analytics: Where Biometric Collection Starts September 12, 2026 12 min. read # BIPA and Video Analytics: Where Biometric Collection Starts Physical Safety Regulatory ComplianceAI Video AnalyticsComputer Vision ![An empty industrial employee entrance corridor with a blank wall-mounted time clock beside a steel door and a small dome camera in the ceiling corner.](/_astro/bipa-video-analytics-biometric-identifier-hero.DQVqytK5_Z2siDuO.webp) Any camera analytics rollout at an Illinois site reaches legal review carrying one question: whether the cameras over the dock will collect biometric data. The Biometric Information Privacy Act answers that with a short list, not a camera type. Where BIPA video analytics risk actually sits is in specific features, and a site that audits for the feature instead of the footage can deploy with a clear answer. ## The problem: the risk gets assigned to the camera BIPA was written around a list. [Section 10 of the Act](https://www.ilga.gov/Legislation/ILCS/Articles?ActID=3004&ChapterID=57) defines a biometric identifier as “a retina or iris scan, fingerprint, voiceprint, or scan of hand or face geometry.” It then lists what is not one: writing samples, signatures, photographs, demographic data, physical descriptions such as height or hair colour, and several medical categories. Biometric information is anything based on one of those identifiers that is used to identify an individual. Two features of that text matter to anyone running cameras. Video is not named in the exclusions — photographs are — so “video is exempt” is not what the statute says. And the thing it does name, a scan of face geometry, is not the footage. It is something computed from the footage. The duties that follow are specific. Under Section 15(b), a private entity may not collect a biometric identifier unless it first informs the person in writing, states the purpose and length of term in writing, and receives a written release. Section 15(a) requires anyone in possession of biometric data to publish a retention schedule and destroy the data when its purpose is satisfied or within three years of the person’s last interaction, whichever comes first. Section 20 sets liquidated damages of 1,000 dollars per negligent violation and 5,000 dollars per intentional or reckless one. The stakes behind those numbers were set out by the Seventh Circuit this April. In _Clay v. Union Pacific Railroad Co._, decided 1 April 2026, a truck driver alleged the railroad scanned his fingerprints about 1,500 times as he entered and left its facilities; on intentional violations, the court noted, that could have meant 7.5 million dollars for one plaintiff. The two companion cases involved fingerprint and hand-geometry time clocks. None of the three involved a camera. That is worth sitting with. All three cases were about the gate and the time clock. A site’s existing BIPA obligations may well sit there, already running, while the legal review argues about analytics on the dock. ## Why the usual approach falls short Two reflexes dominate, and they point in opposite directions. **“Cameras are just video.”** This treats BIPA as irrelevant to anything with a lens. It fails because the statute does not care what captured the image. It cares whether a scan of face geometry was taken from it. A camera running a face recognition feature is squarely within the Act, whatever the camera cost and whatever else it does. **“Any AI on a camera is biometric.”** This treats every analytics feature as face recognition and quietly shelves the deployment for Illinois sites. It fails in the other direction. A detector that reports a pedestrian inside a forklift’s path, or a missing hard hat in a marked zone, produces a location and a label. Refusing it on biometric grounds forfeits the safety case to avoid a risk that feature never created. Between the two sits the vendor questionnaire, which usually contains some version of “is your product BIPA compliant?” A yes to that question tells you nothing, because compliance is a property of how a site collects, notifies, releases and retains, not of a product. What legal review needs is a feature-level answer: which operations in this system produce a face template, can they be turned off, and who holds the output. > BIPA does not regulate the camera. It regulates the template. ## What good looks like: auditing BIPA video analytics risk feature by feature A biometric identifier under Illinois BIPA is one of the items the statute lists by name: a retina or iris scan, a fingerprint, a voiceprint, or a scan of hand or face geometry. Photographs are expressly excluded. For camera analytics, its reach turns on whether a feature computes facial geometry that can identify a person. ### Detection, classification and recognition are different operations The confusion starts with vocabulary. “Computer vision,” “AI analytics” and “facial recognition” get used interchangeably in sales material and in legal memos, and they describe different computations. The clearest technical description comes from NIST. In [NISTIR 8280](https://doi.org/10.6028/NIST.IR.8280), its December 2019 report on demographic effects in face recognition, NIST explains that recognition algorithms run a face detector and then a feature extractor that turns images of a person into what the report calls “a vector of values that relate to the identity of the person.” Recognition then compares two such vectors and scores their similarity. The report contrasts this with classification and estimation, which it describes as single-shot operations on one sample using different machinery. That distinction maps directly onto what a camera analytics system can do. Operation What it outputs Creates a reusable identity record? Object detection A box and a class label: person, forklift, vehicle No Attribute or condition classification A label about that object or scene: hard hat absent, exit obstructed No Face detection The location of a face in the frame Not by itself, but it is the first stage of recognition, so ask what runs next Face recognition or verification A face template compared against stored templates Yes, and this is the scan of face geometry the statute names The table is a technical map, not a legal ruling. Its use is that it turns a vague question — “does this system do biometrics?” — into a concrete one that a vendor can answer and an engineer can check. ### “Can it identify?” is the test courts keep applying The statutory term contains a word courts have taken seriously: identifier. In _Zellmer v. Meta Platforms_, decided on 17 June 2024, the [Ninth Circuit held](https://www.govinfo.gov/content/pkg/USCOURTS-ca9-22-16925/pdf/USCOURTS-ca9-22-16925-0.pdf) that if biometric data “cannot identify an individual, it is not an identifier,” and that face signatures Facebook generated for people who were not users could not identify them on the record before it. The court described that reading as joining other courts that had considered the question. The same opinion records the other side of the line. Discussing an earlier Illinois federal case against Apple, it noted that BIPA applies if face scans could identify a person, even when a company does not use them to. Capability is the test, not intent. Two cautions belong here. _Zellmer_ is a federal appeals court outside Illinois interpreting Illinois law, so it persuades rather than binds Illinois state courts. And “we do not use it to identify anyone” is not a defence if the data could be used that way. ### Control decides who is holding the data The newest development came on 7 August 2026. In _G.T. v. Samsung Electronics America_, the Seventh Circuit considered a phone gallery app that, according to the complaint, grouped photos by facial geometry and built face templates on the user’s own device. The court held that the Act’s key verbs — possess, collect, capture, obtain — all require the company to have some degree of control over the biometric data. Templates that stayed on the user’s device, with no allegation that Samsung could access, modify or use them, fell short; in the court’s words, the case strayed “too far afield from the statute’s heartland.” Read the [G.T. v. Samsung opinion](https://www.govinfo.gov/content/pkg/USCOURTS-ca7-25-01120/pdf/USCOURTS-ca7-25-01120-0.pdf) for the full reasoning. A workplace deployment is a different fact pattern, and it would be a mistake to read _G.T._ as a shield for one. An employer chooses the cameras, configures the analytics and operates the system. If any feature in that system produces face templates, someone controls them. The audit question is who: the site, the vendor’s cloud, or both. ### A five-question feature audit Run these against every analytics product on the camera estate, including anything bundled in camera firmware or the video management system. 1. **Which features extract a face template or any vector used to match the same person across images?** Face recognition, face-based identification of drivers or visitors, watchlists and “find this person” search are the usual candidates. Get the list in writing. 2. **Are those features switched off on Illinois sites, and can they be disabled per camera or per site?** A feature that is licensed but disabled is a configuration fact to document, not an assumption. 3. **If templates exist, where are they stored and who can access them?** This is the control question _G.T._ turned on. Name the servers, the accounts and the vendor’s own access. 4. **If any biometric collection is intended, are the Section 15 steps in place before it starts?** Written notice, a written statement of purpose and term, a written release, and a public retention and destruction schedule. 5. **Does the data-flow diagram show it?** The same document that closes the IT review should name every destination for analytics output. If a template leaves the site, it should appear on that diagram. Then walk to the entrance and check the time clock. In all three cases the Seventh Circuit decided in April, that is where the claim started. ### What this audit does not settle _Clay_ and _G.T._ were decided by a federal court predicting how the Illinois Supreme Court would rule, and the Illinois Supreme Court has the final word on Illinois law. _G.T._ also concerned a consumer device, not a workplace system. The amendment in P.A. 103-769 changed how damages are counted — repeated collection of the same identifier from the same person by the same method is one violation — and the Seventh Circuit held it applies to pending cases. It did not change what counts as an identifier or what a site must do before collecting one. BIPA is also one statute in one state; the rules for camera placement and recording in California are a separate question with separate answers. None of this is legal advice. ## Where Nsightify fits The detections on Nsightify’s two solutions report conditions. On the Physical Safety side they cover PPE compliance, forklift–pedestrian proximity, a blocked exit or obstructed egress path, a person down, and entry into a restricted machine area. On the Perimeter Security side they cover intrusion into a restricted zone, loitering and after-hours zone entry. Each one answers whether a condition exists at a place and a time. Each of those sits on the detection and classification rows of the table above. Nsightify runs on the IP and CCTV cameras a site already operates, with real-time alerts, delivered through Nsightify Cloud or a Zero Trust on-premises deployment in which video stays inside the site’s own network. Two limits belong in the same paragraph. Nsightify is an alerting layer and does not make a site BIPA compliant; notices, releases and retention schedules are steps no product completes on a site’s behalf. And the five audit questions apply to Nsightify exactly as they apply to any other vendor. Detection quality still depends on camera placement, sightlines and lighting. ## Frequently asked questions ### Does video analytics violate BIPA? Not by itself. BIPA regulates biometric identifiers — retina or iris scans, fingerprints, voiceprints, and scans of hand or face geometry — and information based on them that is used to identify someone. Analytics that detects a person, a hard hat or a forklift outputs locations and labels, not a face template. Exposure starts when a feature extracts facial geometry that can identify a person. Confirm your own position with counsel. ### What is the difference between computer vision and facial recognition in BIPA? Facial recognition is one computer vision task among many. NIST describes it as extracting a feature vector that carries identity-related information and comparing it with other vectors, while classification works from a single image without a persistent identity record. BIPA names the scan of face geometry, so a recognition feature is where the statute’s notice and written-release duties come into play. ### How did SB 2979 change video surveillance liability? SB 2979, enacted as Public Act 103-769 and effective 2 August 2024, changed how damages are counted, not what triggers the Act. Repeated collection of the same identifier from the same person by the same method is now a single violation with at most one recovery. The Seventh Circuit held on 1 April 2026 that the change applies to cases pending when it was enacted. ### Can employers use AI cameras in Illinois? Yes. Illinois does not prohibit camera analytics. BIPA requires written notice, a stated purpose and retention term, and a signed written release before a private entity collects a biometric identifier, plus a public retention and destruction schedule. Detections that never create a face template do not collect one. Features that do — face-based identification in particular — need those steps first. * * * ## What to put in front of legal review this week Replace the question “is this system biometric?” with a one-page feature inventory. List every analytics product on the Illinois camera estate, mark each feature against the four operations in the table, and attach the vendor’s written answer to the five audit questions. Add the biometric time clock and any fingerprint readers at the gate while you are at it, because they are the more likely source of an existing obligation. That page gives counsel something to rule on. It also belongs next to the [stream inventory and data-flow diagram IT asks for before approving video analytics](/nsights/video-surveillance-network-requirements-it-approval/), and it answers several of the [video analytics RFP questions vendors cannot bluff](/nsights/video-analytics-rfp-vendor-questions/). Sites operating in more than one state should run the same exercise against [California’s limits on where workplace cameras can go](/nsights/california-workplace-surveillance-camera-placement/), which turn on rooms rather than templates. If you want to see which safety conditions your existing Illinois cameras can report, [talk to us about Physical Safety](/contact/?interest=physical-safety). More on this from Nsightify: [PPE detection and hazard-zone monitoring](/solutions/physical-safety/). [Previous](/nsights/california-workplace-surveillance-camera-placement/) [California Workplace Surveillance Laws: Where Cameras Cannot Go](/nsights/california-workplace-surveillance-camera-placement/) [Next](/nsights/how-accurate-is-ai-video-analytics-testing-standard/) [How Accurate Is AI Video Analytics? Test It, Don't Trust It](/nsights/how-accurate-is-ai-video-analytics-testing-standard/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## California AI Workplace Law SB 947: A Human Must Corroborate URL: https://nsightify.com/nsights/california-ai-workplace-law-sb-947/ The California AI workplace law, SB 947, now signed, requires a human to corroborate AI-based discipline from 1 July 2027. What it means for camera alerts. [Home](/) [Nsights](/nsights/) California AI Workplace Law SB 947: A Human Must Corroborate August 17, 2026 Updated October 1, 2026 15 min. read # California AI Workplace Law SB 947: A Human Must Corroborate Physical Safety Regulatory ComplianceAI Video AnalyticsPPE Compliance ![An empty shift supervisor's desk at the edge of a distribution center floor, with a two-way radio on its charger and a clipboard, in flat daylight.](/_astro/california-ai-workplace-law-sb-947-hero.DpbRGsM5_Z236IUk.webp) A PPE alert lands in a supervisor’s queue at 10:14 on a Tuesday. Who reads it, what they write down, and whether anything happens to the person in the frame is what the California AI workplace law, SB 947, will govern from 1 July 2027. The detection is not the regulated act — the decision after it is. ## SB 947 is signed: its requirements start on 1 July 2027 SB 947, which its author calls the No Robo Bosses Act of 2026, adds a new part to the Labor Code — Part 5.5.5, commencing with section 1520 — governing how employers use automated decision systems in decisions about their employees. Senator Jerry McNerney introduced it on 2 February 2026. It was amended several times, most recently on the Assembly floor on 21 August 2026. The [bill’s history](https://leginfo.legislature.ca.gov/faces/billHistoryClient.xhtml?bill_id=202520260SB947) shows how close it came. It passed the Senate in May, went to the Assembly, and was placed on the Appropriations suspense file on 5 August 2026 — normally where a bill with a fiscal note goes to die quietly. On 13 August the Assembly suspended its own Rule 63, the committee reported it out on a 10–4 vote, and it was read a second time and ordered to third reading. It was amended on the floor on 21 August, passed the Assembly 53 to 14 on 30 August, and the Senate concurred 28 to 10 on 31 August. The Governor signed it on 30 September 2026. As a non-urgency statute it takes effect on 1 January 2027 under article IV, section 8(c) of the California Constitution, and section 1526.7 delays its operation to 1 July 2027. The signature is worth reading against what came before. Its predecessor, SB 7, passed both houses in 2025 and was vetoed on 13 October 2025. Governor Newsom’s message called the restrictions “overly broad” and the notification requirements unfocused, and said forthcoming California Privacy Protection Agency regulations should be assessed before new legislation. [Fisher Phillips](https://www.fisherphillips.com/en/insights/insights/california-governor-vetoes-no-robo-bosses-act) and several other employment firms published the same reading within days. SB 947 is the narrower answer to that veto. The most consequential change: it drops SB 7’s 30-day pre-use notice, which drew most of the fire, and keeps only the post-use notice. _Current as of 1 October 2026. Governor Newsom approved SB 947 on 30 September 2026 (Chapter 859, Statutes of 2026). It takes effect on 1 January 2027, and its requirements become operative on 1 July 2027. Nothing below is legal advice._ ## What California’s AI workplace law counts as an automated decision system An automated decision system, in SB 947’s terms, is any computational process derived from machine learning, statistical modeling, data analytics or artificial intelligence that issues simplified output, including a score, a classification or a recommendation, used to assist or replace human discretionary decisionmaking, and that materially impacts natural persons. Spam filters, firewalls, antivirus software, calculators and databases are carved out. Read that definition three times and the operative words separate out. **Simplified output.** Not a report, not a dashboard, not a video clip. A score, a classification, a recommendation. A detection that returns “no hard hat, zone 4, 10:14” is a classification. **Assist or replace.** This is the phrase that catches systems nobody thinks of as decision-making. The process does not have to decide anything. It only has to feed a discretionary decision that a person then makes. **Materially impacts natural persons.** The output has to land on somebody. The [bill text](https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260SB947) also defines employee data broadly — any information that identifies, relates to or describes an employee, however it was collected, inferred or obtained — and counts labor contractors and farm labor contractors as employers. A yard run through a labor contractor is not outside this. The law sets no minimum employer size, covers public employers as well as private ones, and carries a civil penalty of $500 per violation, enforced by the Labor Commissioner or a public prosecutor. So the question for anyone running safety or security analytics is not whether a camera detection is an automated decision system in the abstract. It is narrower and more answerable: is the output of that detection being used to assist a discretionary decision that lands on a worker? For most sites, the honest answer is that nobody has ever written it down. ## A detection is not a decision The law’s core duties do not attach to observing. They attach to relying. From 1 July 2027, it prohibits an employer from relying solely on an automated decision system when making a disciplinary or termination decision. And where an employer primarily relies on such a system’s output for one of those decisions, the employer must direct a human to corroborate the decision, using the data behind the output or other supporting information, such as supervisory evaluations, personnel files, work product, peer reviews or witness interviews. Output that cannot be corroborated cannot be used for the decision. > A camera can report that a hard hat was missing at 10:14. It cannot report why. Note the threshold word. Not “uses”. Not “consults”. _Primarily relies._ A detection that starts a conversation is in a different position from a detection that ends one, and the difference is entirely in what the organization does with it. There is a second set of provisions worth knowing about, because they rule out uses rather than workflows. SB 947 bars using an automated decision system to infer an employee’s protected status, or to predict and take adverse action against a worker for exercising their legal rights. Emotion inference sits in a different law signed the same day: [California’s emotion AI workplace law, AB 1883](/nsights/california-emotion-ai-workplace-law-ab-1883/), bars an employer from using a workplace surveillance tool that runs artificial intelligence to infer an employee’s emotional state. Either is a different thing from detecting whether high-visibility clothing is present in a frame. Instrument Status as of 1 October 2026 What it reaches SB 947 (Chapter 859, Statutes of 2026) Signed 30 September 2026; operative 1 July 2027 Reliance on an automated decision system for discipline or termination CCPA automated decisionmaking regulations (CPPA) In effect since 1 January 2026, with the ADMT obligations phasing in from 1 January 2027 Use of automated decisionmaking technology for a “significant decision”, which the rules define to include employment SB 7 — the 2025 predecessor Vetoed 13 October 2025 Nothing. Would have required notice 30 days _before_ use, as well as after The middle row is the one people miss. The [California Privacy Protection Agency’s regulations](https://cppa.ca.gov/regulations/ccpa_updates.html) are not a proposal. They took effect on 1 January 2026, they treat employment decisions as significant decisions, and they turn on a concept called human involvement: a person who knows how to interpret the output, reviews and analyzes it, and has the authority to make or change the decision. SB 947 keeps employers the CCPA covers subject to those rules. Two different California instruments, arriving at the same requirement from opposite directions. ## Why “a human looks at it” is not the answer Every operations manager asked about this says the same thing, and they are not being evasive. Nobody has automated discipline. A supervisor reads the alert. A human is in the loop. On most floors that is literally true. It is also not what either instrument asks for. Three gaps open up under pressure. **The duty is not satisfied by presence.** A supervisor who opens the alert, agrees with it, and issues a written warning has read the system’s output, nothing more. The law asks a human to corroborate the decision against the data behind the output or other supporting information, and bars using output that cannot be corroborated. A second, independent source is good practice, not the statutory minimum. **Nothing survives the review.** The obligations described here all produce documents: corroborating information, a written notice to the worker, and a description of the worker’s data if they ask for one. A workflow that runs through a group chat and a supervisor’s memory produces none of them. Six months later there is no way to demonstrate what was actually done, which is the same problem as not having done it. **The alert queue has no owner.** Ask who is authorized to act on a forklift-proximity alert on the night shift and most sites cannot answer with one name. If nobody owns the decision, nobody is accountable for how it was made — and the record of it is whatever each shift improvised. Be fair about the timing: none of this binds until 1 July 2027. But the CPPA’s automated decisionmaking obligations apply from 1 January 2027, and the workflow question is identical under both. Building the review step now costs a policy document and a form. Building it after a claim, a citation or a regulator’s request costs considerably more. ## What the workflow needs to look like None of this is a technology change. Every item below is a decision about who does what, written down before the alert arrives. ### Separate the detection record from the decision record A detection log and a personnel file are two different systems with two different retention profiles, two different audiences, and two different legal exposures. When they are the same spreadsheet, every hazard observation becomes a potential piece of employment evidence, and the whole log inherits the sensitivity of its worst row. Keep the hazard record about conditions: zone, time, what was observed, what was done to fix it. Anything that becomes a personnel matter gets escalated into a separate file, deliberately, by a named person. ### Name who is allowed to act on an alert One name per detection type per shift. Not a role, not a distribution list — a person, and a named deputy. The point is not bureaucracy. It is that “primarily relies” is a factual question about a specific decision, and a decision with no author cannot be defended. ### Make corroboration produce an artifact Beyond the statutory minimum, good corroboration for a camera alert is something the camera did not produce. A supervisor’s own observation, written the same day. The shift’s PPE issue log. The toolbox talk record. The worker’s own account of what happened, recorded before any decision. The test is simple and worth applying to your current process: could you hand the file to someone who has not seen the clip, and would they reach the same conclusion? If the answer is no, the clip is the decision and the review was a formality. ### Draft the notice before you need one The post-use notice attaches at the moment the employer informs the worker of the decision — not a week later. It has to be a separate, plain-language written communication, and section 1524 fixes its four contents: that the employer primarily relied on an automated decision system, that a human reviewed the decision and corroborated the output, a person to contact about the decision and about the worker’s right to a description of their data, and that retaliation is prohibited. That is a template, and a template written under time pressure on the day of a termination is a template written badly. ### Decide the data-access question now SB 947 lets an employee request, and requires the employer to provide, a meaningful, objective description of the employee’s own data the system used, where an automated decision system was primarily used for a disciplinary or termination decision. Other people’s data in that description has to be anonymized. Nobody can describe data they no longer hold, so this is a retention question as much as a privacy right. The resolution is not more storage. It is deciding, in advance, which detection metadata is retained long enough to describe and substantiate a decision and which video is not retained at all — a design choice, made once, rather than a default inherited from a recorder’s disk size. ### Keep the safety case and the performance case apart This item does the most work and needs the least technology. Route safety detections to hazard removal, not to individual performance scoring. A blocked-exit alert should reach whoever can move the pallet. A forklift-proximity alert should reach whoever can re-mark the aisle or change the traffic pattern. That is better safety practice on its own terms — it is what [real-time physical safety monitoring on cameras you own](/nsights/physical-safety-monitoring-real-time-ai/) is actually for, and it is the posture an inspector rewards under [OSHA’s warehouse national emphasis program, which now runs to 2031](/nsights/osha-warehouse-emphasis-program-2026/). It also happens to keep the great majority of safety alerting outside the reach of a statute about disciplinary decisions, because no disciplinary decision is being made. ## What an Nsightify detection reports, and what it leaves to people An Nsightify detection reports that a condition exists at a place and a time. PPE compliance, forklift–pedestrian proximity, restricted-area dwell, blocked exit, person down — each produces a real-time alert that a state was observed, on a site’s existing cameras. It issues no score about a person, ranks nobody, and takes no action. The discretionary decision stays where SB 947 puts it: with a human who has to check. The limit has to be stated plainly. Nothing in that design stops a supervisor from treating an alert as a verdict. The moment a detection becomes the primary basis for a written warning, the employer is inside the workflow this law regulates, and no vendor architecture changes that. The control is the employer’s written policy and the sequence people actually follow. Using Nsightify resolves no obligation under SB 947, and nothing here is legal advice. The technical constraints matter for the same reason. Camera-based detection depends on sightlines, lighting, and where a camera was mounted — which was almost never for analytics. Occlusion behind racking is real. A missing-vest classification is a statement about pixels in a frame, which is precisely why it belongs in a hazard workflow as an observation, and not in a personnel file as a finding. Working out [what your existing cameras can already tell you](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) is a separate question from deciding who is allowed to act on it. ## Frequently asked questions ### Can AI fire your employees in California? Not on its own, from 1 July 2027. SB 947 (Chapter 859, Statutes of 2026) prohibits an employer from relying solely on an automated decision system for a disciplinary or termination decision. Where an employer primarily relies on such a system’s output for one, a human must corroborate the decision, and output that cannot be corroborated cannot be used. Confirm your own position with counsel. ### Does California require human review of AI decisions? SB 947 does, from 1 July 2027. It bars sole reliance on an automated decision system for discipline or termination, and requires a human to corroborate any such decision that primarily relies on the system’s output. Separately, the California Privacy Protection Agency’s automated decisionmaking regulations, in effect since 1 January 2026, treat human involvement as what takes a technology outside their scope. Confirm your own obligations with counsel. ### What counts as an automated decision system? Under SB 947, any computational process derived from machine learning, statistical modeling, data analytics or artificial intelligence that issues simplified output, including a score, a classification or a recommendation, used to assist or replace human discretionary decisionmaking, and that materially impacts people. The word assist carries the weight: a system does not have to decide anything to be inside the definition. ### Can my employer use camera footage to discipline me? Nothing in SB 947 prohibits that. What it changes, from 1 July 2027, is the sequence. Where an automated system’s output is the primary basis for a disciplinary decision, a human must corroborate the decision, and the employer must state in a written notice, at the time it informs the worker of the decision, that it primarily relied on such a system. ### Do I have to tell employees about workplace monitoring in California? SB 947 is not the instrument that answers that. Its notice duty is post-use: it attaches at the moment an employer informs a worker of a disciplinary or termination decision, not at the moment monitoring begins. Notice obligations for the collection of employee personal information sit under the CCPA and its regulations instead. Confirm your own position with counsel. * * * ## One page to write before July 2027 None of the above needs to wait for 1 July 2027. Take one detection type you already run — PPE is the usual candidate — and answer four questions on one page. Who receives the alert on each shift. Who is authorized to escalate it into a personnel matter. What corroboration that person must gather before they do. How long the underlying record is kept, and where. That page is the difference between an observation and a decision, and it is the artifact both California instruments are ultimately asking to see. If the honest answer to any of the four is “it depends who is on”, the gap is in the workflow rather than in the statute — and it is the same gap that makes an incident review two years later impossible to run. Writing that page can raise questions about how camera alerts reach people on each shift, or what an alert actually records. [Put them to the Nsightify team](/contact/?interest=question). **Keep reading:** * [OSHA’s warehouse national emphasis program now runs to 2031](/nsights/osha-warehouse-emphasis-program-2026/) * [Real-time physical safety monitoring on cameras you own](/nsights/physical-safety-monitoring-real-time-ai/) More on this from Nsightify: [PPE detection and hazard-zone monitoring](/solutions/physical-safety/). [Previous](/nsights/ndaa-compliant-cameras-fcc-import-ban/) [NDAA Compliant Cameras: You May Not Need to Replace Yours](/nsights/ndaa-compliant-cameras-fcc-import-ban/) [Next](/nsights/video-analytics-rfp-vendor-questions/) [Video Analytics RFP Questions: 9 Tests Vendors Cannot Bluff](/nsights/video-analytics-rfp-vendor-questions/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## California Emotion AI Law: The Definition Is Wider Than the Ban URL: https://nsightify.com/nsights/california-emotion-ai-workplace-law-ab-1883/ The California emotion AI workplace law, signed in September 2026, bans two AI uses from 2027 and defines surveillance tools more broadly. What AB 1883 covers. [Home](/) [Nsights](/nsights/) California Emotion AI Law: The Definition Is Wider Than the Ban September 21, 2026 Updated October 1, 2026 11 min. read # California Emotion AI Law: The Definition Is Wider Than the Ban Physical Safety Regulatory ComplianceAI Video AnalyticsComputer Vision ![A ceiling-mounted security camera on a steel roof truss inside a distribution warehouse, with racking and concrete floor out of focus below in flat daylight.](/_astro/california-emotion-ai-workplace-law-ab-1883-hero.BiT2xo-X_PUGfl.webp) A camera that flags a missing hard hat and a system that scores whether a worker looks frustrated are, under California’s emotion AI workplace law, the same category of thing. The definition reaches both. The prohibition reaches only one of them, and the gap between those two sentences is where every operations manager in the state now has to do some reading. ## The California emotion AI workplace law starts with a definition AB 1883, signed on 30 September 2026 as Chapter 853, Statutes of 2026, adds a new part to the Labor Code — Part 5.8, commencing with section 1580 — governing workplace surveillance tools. Assemblymember Bryan is the author. The new part runs to four sections, 1580 to 1583, in the [enrolled text](https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260AB1883) the Governor signed: definitions, a prohibition, enforcement, and a narrow exemption. Most coverage of the law leads with the prohibition, which is the newsworthy part. The operational problem starts one section earlier, in the definitions. A workplace surveillance tool, as section 1580 defines it, is any system, application, instrument or device that collects or facilitates the collection of employee data, activities, communications, actions, biometrics or behaviors — gathered, in the statute’s words, “by means other than direct observation by a person.” That last clause is the whole reach of the definition. > The definition describes how information is gathered, not what it is gathered for. Read it against your own site. A camera running PPE detection collects data about employee actions by means other than a person watching. So does a badge reader, a time-tracking application, a telematics unit in a yard truck. The definition is not a description of invasive technology. It is a description of automation. _Status, current as of 1 October 2026: AB 1883 was enrolled on 3 September, presented to the Governor on 10 September, and approved and chaptered on 30 September 2026 as Chapter 853, Statutes of 2026. It takes effect on 1 January 2027. The [bill’s status page](https://leginfo.legislature.ca.gov/faces/billStatusClient.xhtml?bill_id=202520260AB1883) is the record. Nothing below is legal advice._ ## Three readings of AB 1883, and where each one fails Three readings of this law are circulating, and each one fails in a different way. **The headline reading.** “California bans emotion AI at work” is accurate and tells an employer nothing actionable. It does not say what an emotion AI system is, whether the reader owns one, or what happens to the equipment that is already installed. **The client-alert reading.** Employment firms have covered the bill competently, and they are writing for general counsel — enumerating what passed, what it prohibits, what the penalty is. That is the right document for a lawyer and the wrong one for a site manager, because it stops at the prohibition and never tests the definition against real equipment. **The assumption reading**, which is the dangerous one. It goes: we run safety cameras, safety is obviously fine, nothing here concerns us. The conclusion may well be right. The reasoning is not, because it skips the two sections that decide the question, and an employer who cannot say which section protects them cannot answer an inspector, an auditor or their own safety committee. The law is short. Reading it in order takes ten minutes and produces a defensible position instead of an assumed one. ## How to read AB 1883’s three moving parts in order AB 1883 has a definition, a prohibition and an exception, and they do different jobs. Taken in sequence they answer the question; taken out of sequence they produce the assumption reading above. Section What it establishes What it means on your floor 1580 Defines a workplace surveillance tool by collection method, not by purpose Most automated systems that observe employees are inside the definition, including safety cameras 1581(a) Prohibits two uses of such a tool where it uses artificial intelligence: recognizing, inferring or predicting an individual’s emotional state, and collecting neural data The question is what your system infers, not what it is called 1581(b) States the section does not prohibit using such a tool to ensure safety, or where it does not do the two things in 1581(a) Two separate routes out of the prohibition, and they are not the same route ### The prohibition targets two uses, not cameras Section 1581(a) does not regulate a device class. It regulates two uses: one inference and one kind of collection. The first is recognizing, or making inferences or predictions about, an individual’s emotional state. The second is collecting neural data, which section 1580 defines as information generated by measuring the activity of an employee’s central or peripheral nervous system that is not inferred from nonneural information. That framing matters more than it first appears. A detection that reports a physical fact in a scene — a person is in a restricted zone, required head protection is absent, a forklift and a pedestrian are converging — is making a claim about the world. A system that reports that a worker seems stressed is making a claim about a person’s interior state. The statute draws its line there, and it is a line a buyer can actually test a product against. ### The exception has two limbs, and they are different Section 1581(b) does two things in one sentence. It says the section does not prohibit an employer from using a workplace surveillance tool to ensure safety. It also says the section does not prohibit using one that does not meet the specifications in subdivision (a). Those are not the same argument. The first is a purpose-based carve-out and its boundaries are not defined in the text — the statute does not say what ensuring safety covers, who decides, or what happens to a system used for safety on Monday and productivity on Tuesday. The second is simpler and stronger: a tool that never infers emotional state and never collects neural data is outside subdivision (a) on its own terms, with no reliance on purpose at all. For most safety deployments the second limb is the one worth understanding, because it turns on a verifiable property of the product rather than on a characterization of intent. ### The term the statute uses but does not define Section 1580 is a definition section. It defines artificial intelligence, employ, employee, employer, neural data, public prosecutor and workplace surveillance tool. It does not define emotional state. That is an observation about the drafting, not a prediction about what follows from it. It is worth noting because the prohibition’s scope depends on a term the law leaves to ordinary meaning, and because capabilities such as fatigue scoring, gait analysis or behavioral risk prediction would sit at unclear distances from it. An employer’s exposure under this section is a function of what their systems infer, so the sensible response is to find out precisely what each one claims to infer. ### Five questions to put to a vendor before 1 January 2027 Ask for written answers. A vendor who cannot answer these in writing has told you something. 1. **Does the system infer, score or predict any emotional, psychological or affective state of a person?** Not “do you market it that way” — does any model in the pipeline output such a value. 2. **Does it collect neural data** as section 1580 defines it, from any sensor, in any configuration. 3. **What exactly does each detection assert?** A physical event in a scene, or a characterization of a person. 4. **Are any inference features switched off by default but available?** A capability you have not enabled is still a capability you have bought. 5. **What is the stated purpose of each detection in our deployment**, and is that purpose recorded anywhere you could show a third party. ## What Nsightify’s detections report, and what they don’t Nsightify’s detections report physical events in a scene. Whether required PPE is present, whether a person has entered a restricted machine area, whether a forklift and a pedestrian are converging, whether a person is down, whether an exit is blocked. Each one is a statement about what is happening in the camera’s field of view. None of them infers an employee’s emotional state, and none measures nervous-system activity. The two capabilities section 1581(a) prohibits are not capabilities Nsightify offers — which is a statement about what the product does, not a legal conclusion about any particular deployment. Whether a given installation sits inside or outside the statute is a question for the employer and their counsel, and this article does not answer it. The limits are worth stating in the same breath. A detection depends on camera placement, sightlines, lighting and weather, and a record of what a camera could see is not a record of the whole site. Detection is also not prevention: an alert tells a supervisor something is happening, and what follows is a human decision. Where that decision is discipline or termination resting mainly on automated output, [California’s AI workplace law, SB 947](/nsights/california-ai-workplace-law-sb-947/), requires a human to corroborate it. SB 947 was signed on 30 September 2026 as Chapter 859 and becomes operative on 1 July 2027. Nsightify runs on a site’s existing cameras, with real-time alerting, and publishes no accuracy or incident-reduction figure for any detection. ## Frequently asked questions ### Is emotion AI legal in California? Not on employees, from 1 January 2027, with two exceptions. AB 1883, signed on 30 September 2026, bars an employer from using a workplace surveillance tool that uses artificial intelligence to recognize, infer or predict an individual’s emotional state, or to collect neural data. Use to ensure safety is not prohibited, and federally required aerospace and defense work is exempt. The Labor Commissioner and public prosecutors enforce it, with penalties of up to $500 per violation. ### What counts as a workplace surveillance tool? Under AB 1883’s definition, any system, application, instrument or device that collects or facilitates the collection of employee data, activities, communications, actions, biometrics or behaviors by means other than direct observation by a person. The phrase is deliberately wide. A camera running safety detections, a time-tracking system and a geolocation tool all sit inside it, because the definition describes how information is gathered rather than what it is used for. ### Does AB 1883 apply to security cameras? A camera system falls inside the law’s definition of a workplace surveillance tool. Falling inside the definition is not the same as being prohibited. Section 1581 bans two specific uses — emotional state inference and neural data collection — and then states that the section does not prohibit using such a tool to ensure safety, or where the tool does not do the two prohibited things. Confirm your own position with counsel. ### When does the California AI workplace law take effect? On 1 January 2027. AB 1883 was approved by the Governor on 30 September 2026 and chaptered as Chapter 853, Statutes of 2026. It has no urgency clause and no delayed operative date, so article IV, section 8(c) of the California Constitution puts its effective date on the 1 January after a 90-day period from enactment. ### What is the penalty for violating AB 1883? AB 1883 lets the Labor Commissioner or a public prosecutor enforce it, with a penalty of up to $500 per violation, paid either as a civil penalty or as a statutory penalty to the employee, not both. Injunctive relief, punitive damages and attorney fees are also available. Work on aircraft for the national airspace, or on national security, military, space or defense products, is exempt where the tool is reasonably necessary to comply with federal law or a binding federal contract. * * * ## What to do about this before January Take the inventory before you take a position. List every automated system on site that observes employees, and next to each one write what it infers — a physical event, or something about a person. Most lists come back shorter and duller than expected, which is the useful outcome. Then check the dates yourself. AB 1883 was signed on 30 September 2026, the last day [Article IV, section 10 of the California Constitution](https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?lawCode=CONS§ionNum=SEC.%2010.&article=IV) allowed, and takes effect on 1 January 2027. Read the status page rather than the headline. The same inventory answers more than one question. It is the list that shows [where California workplace surveillance laws say cameras cannot go](/nsights/california-workplace-surveillance-camera-placement/), including under AB 1331, signed the same day, which uses nearly the same definition to bar monitoring employees in workplace bathrooms. It tells you [where biometric collection actually starts under BIPA](/nsights/bipa-video-analytics-biometric-identifier/) if you operate in Illinois too, and [how employee monitoring laws in other states treat cameras](/nsights/employee-monitoring-laws-by-state-cameras/). Where a line on that inventory leaves you unsure what a detection actually measures, [ask the Nsightify team](/contact/?interest=question). More on this from Nsightify: [PPE detection and hazard-zone monitoring](/solutions/physical-safety/). [Previous](/nsights/how-accurate-is-ai-video-analytics-testing-standard/) [How Accurate Is AI Video Analytics? Test It, Don't Trust It](/nsights/how-accurate-is-ai-video-analytics-testing-standard/) [Next](/nsights/ppe-detection-what-cameras-cannot-verify/) [PPE Detection: What a Camera Verifies, and What It Cannot](/nsights/ppe-detection-what-cameras-cannot-verify/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## California Workplace Surveillance Laws: Where Cameras Cannot Go URL: https://nsightify.com/nsights/california-workplace-surveillance-camera-placement/ California workplace surveillance laws bar recording in three named rooms, and AB 1331 bars bathroom monitoring from 2027. How to audit your camera map. [Home](/) [Nsights](/nsights/) California Workplace Surveillance Laws: Where Cameras Cannot Go September 8, 2026 Updated October 1, 2026 15 min. read # California Workplace Surveillance Laws: Where Cameras Cannot Go Physical Safety Regulatory ComplianceCamera IntegrationAI Video Analytics ![A long internal corridor in an industrial facility, with a wall-mounted security camera in the foreground angled away down the corridor past two plain closed doors.](/_astro/california-workplace-surveillance-camera-placement-hero.DWkt5kaB_2ppSn3.webp) A camera map is drawn by whoever mounted the cameras, usually years ago, for reasons nobody wrote down. California workplace surveillance laws name three rooms where a recording of an employee is prohibited outright — and a law signed on 30 September 2026 puts a per-violation penalty, from 1 January 2027, behind a narrower version of the same rule. Placement stops being a facilities decision at that point and becomes something a site has to be able to defend. ## What California workplace surveillance laws already prohibit Whatever the new law below adds, the operative rule has been on the books since 1 January 1999 and is short enough to read in full. [Labor Code section 435](https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?sectionNum=435.&lawCode=LAB) prohibits an employer from causing an audio or video recording to be made of an employee in a **restroom**, a **locker room**, or a **room designated by the employer for changing clothes**, unless a court order authorizes it. Subdivision (b) adds a second, separate prohibition that is routinely missed: a recording made in violation of the section may not be used by the employer **for any purpose**. It applies to private and public employers alike, excepting the federal government. A violation is an infraction. Two things in that paragraph do more work than they appear to. **It names rooms, not expectations.** Most writing on workplace privacy reasons from a reasonable expectation of privacy, which is a standard applied after a dispute exists. Section 435 does not ask a site to predict how a court would weigh anything. It names three rooms. **The bar on use is independent of the bar on recording.** If footage of one of those rooms exists — because a camera was aimed badly, because a changing area was designated after the camera went in, because a door was propped open — the employer cannot use it. Not in an investigation, not in a termination, not as evidence of theft. The recording being useful does not make it usable. ## AB 1331 is now law: bathroom monitoring is banned from 2027 AB 1331, authored by Assemblymember Elhawary, adds a new part to the Labor Code — Part 5.8, commencing with section 1560 — governing workplace surveillance tools. The [bill’s history](https://leginfo.legislature.ca.gov/faces/billHistoryClient.xhtml?bill_id=202520260AB1331) shows a bill that nearly died and then moved quickly. It went to the Senate inactive file on 13 September 2025 and sat there for eleven months. It was pulled off that file on 21 August 2026, amended on 21 and 28 August, and on 31 August — the last day either house could pass bills — the Senate passed it 28 to 5 and the Assembly concurred in the Senate amendments 61 to 7. Its enrolled text is dated 4 September 2026, and it was presented to the Governor on 14 September. Article IV, section 10(b) of the California Constitution gave the Governor until 30 September to act on it. He signed it on **30 September 2026**, and it was chaptered the same day as Chapter 851, Statutes of 2026. It has no urgency clause and no operative date of its own, so article IV, section 8(c) sets the effective date: a statute passed at a regular session takes effect on the January 1 following a 90-day period from enactment, which puts it at **1 January 2027**. _Current as of 1 October 2026. AB 1331 was signed on 30 September 2026 (Chapter 851, Statutes of 2026) and takes effect on 1 January 2027. Nothing below is legal advice._ ## What AB 1331 adds that Labor Code 435 does not reach A workplace surveillance tool, under new Labor Code section 1560(f), is a system, application, instrument or device that collects, or facilitates the collection of, employee data, activities, communications, actions, biometrics or behaviors by means other than direct observation by a person. Smoke and carbon monoxide detectors are excluded, and so are weapon detection systems that automatically screen a person’s body. A fixed camera running analytics is squarely inside that definition. [California’s emotion AI law, AB 1883](/nsights/california-emotion-ai-workplace-law-ab-1883/), signed the same day, defines the term in nearly the same words and bars two uses of a tool that runs artificial intelligence: inferring a worker’s emotional state and collecting neural data. The two instruments are easy to conflate and do different work. Labor Code 435 (law since 1999) AB 1331 (signed 30 September 2026; in force 1 January 2027) Conduct reached Causing an audio or video **recording** to be made Using a surveillance tool to **monitor or surveil** Places covered Restroom, locker room, changing room Bathroom only Employee right None stated Right to leave the tool behind when entering a bathroom Enforcement Infraction Labor Commissioner citation or civil action; public prosecutors may also act Penalty Infraction penalty Up to 500 dollars per violation Notice duty None **None** — the law adds no notification requirement > Labor Code 435 reaches the recording. AB 1331 reaches the watching. That distinction is the whole delta. A camera that can see into a workplace bathroom, streams to a monitor and records nothing sits outside the plain words of section 435, which is about causing a recording to be made. From 1 January 2027, it sits inside AB 1331’s prohibition on monitoring. AB 1331 also carves out a narrow exception on the other side: an employer may still check a surveillance tool for the one-time entry and exit of bathrooms, so a device that logs a door opening is treated differently from one that watches what happens beyond it. The right to leave a tool behind has narrow limits. It does not apply when federal or state law requires an employee to stay available during meal or rest periods. And an employer is not in violation when an employee voluntarily takes a tool into a bathroom, or takes one the employer requires for identification or safety, such as a badge or personal alarm, or to open a locked bathroom. That second case holds only if the tool neither detects nor records audio or video and has no artificial intelligence embedded or attached. Two things the law does **not** do are as important as what it does. It restricts bathrooms and nothing else — no break room provision, no locker room provision, no parking or personal vehicle provision. And it imposes no notice duty at all. A reader who takes the headline and assumes a general workplace surveillance statute has arrived will act on a rule that is not there. ## California has no 14-day notice rule for workplace cameras Search for this and page one gives two kinds of page, neither of which produces an answer an operator can act on. **The employment-firm summary** explains reasonable expectation of privacy at length, correctly, and never names a room. It is written to establish that a claim might exist, which is a different job from telling a security manager whether the camera above the locker aisle is a problem. **The state-by-state comparison table** names California and then repeats a claim that will not survive contact with the statute: that California employers must give 14 days’ written notice before monitoring. No California statute imposes that. The nearest real duty runs the other way, and only for employers the California Consumer Privacy Act (CCPA) covers: for-profit businesses that meet one of its size or data thresholds. For them, [Civil Code section 1798.100](https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?sectionNum=1798.100.&lawCode=CIV) requires disclosing the categories of personal information collected, the purposes, and how long each category is kept, **at or before the point of collection**. That is a content requirement with no waiting period attached. A bill that would have imposed broader duties on workplace surveillance tools, AB 1221, died on 31 January 2026. Getting this wrong is expensive in both directions. A site that believes it owes 14 days’ notice delays a safety deployment for a rule that does not exist. A site that concludes the notice question is therefore fictional can skip a disclosure it does owe. ## How to audit a camera map against California law An audit here is not a legal opinion. It is an inventory question, and the site already has most of the answer written down somewhere. ### The four questions to ask about each camera Run these against every device on the map, one row per camera: 1. **What is in the frame, including at the edges?** Not what the camera is for — what it can see. A dock camera whose field of view clips the corridor outside a changing room is a section 435 question even though nobody aimed it there. 2. **Does it record, monitor, or both?** Section 435 is about recording. AB 1331 is about monitoring. From 1 January 2027, a live-only feed and a recorded one sit under different rules. 3. **Does it capture audio?** If yes, the question changes statute. [Penal Code section 632](https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?sectionNum=632.&lawCode=PEN) requires the consent of all parties to record a confidential communication, with a fine of up to 2,500 dollars per violation and up to 10,000 after a prior conviction. Check every camera with a built-in microphone. 4. **Has the room changed since the camera went in?** A storage bay designated as a changing area, a corridor converted to a locker aisle. The camera did not move; the statute’s coverage did. ### The areas that are settled, and the areas that are argued The settled list is short: the three rooms named in section 435, plus workplace bathrooms, which AB 1331 covers for monitoring from 1 January 2027. Everything else is decided on facts. The break room is the clearest example, and the most asked about. No California statute names it. It is not a restroom, a locker room, or a room designated for changing clothes, so section 435 does not reach it, and AB 1331 does not either. Whether a camera there is defensible turns on whether employees reasonably expected privacy in that space — which depends on signage, on layout, on whether the room is enclosed, and on what the camera is being used for. That is a fact question, and a site that documents its answer in advance is in a better position than one that improvises it after a complaint. The useful move is to sort the map into three lists rather than two: prohibited, defensible with a documented reason, and unexamined. The third list is where the risk actually lives. ### Notice: write the disclosure, skip the waiting period The CCPA duty is a content duty. It asks what categories of personal information are collected, why, and for how long each is kept — delivered at or before collection. For a camera estate that usually means a posted notice at entrances and a line in the employee handbook that matches what the system actually does, including retention. Several other states handle notice through employee-monitoring statutes that sort cameras by purpose; [how employee monitoring laws by state treat cameras](/nsights/employee-monitoring-laws-by-state-cameras/) compares four of them. The retention half is where these documents most often go wrong, because the honest answer is frequently whatever the recorder holds before it overwrites, which is a storage specification rather than a policy. ## Camera exclusion and zone masking, and what they don’t change Analytics inherits the camera map it is given. Nsightify runs on a site’s existing cameras, so the placement audit above is a prerequisite to a deployment rather than an output of one. The useful part of that constraint is that the places California puts off-limits are places no safety detection needs. PPE compliance, forklift–pedestrian proximity, blocked exit and obstructed egress, person down, restricted-area dwell — every one lives on a floor, a dock, an aisle or a yard. A restroom, a locker room and a changing room contribute nothing to any of them. Where the map and the statute do not line up neatly, the configuration can be made to. A camera can be excluded from analysis entirely, and a zone inside a camera’s field of view can be masked so it is never analyzed — the dock camera whose frame clips the corridor outside a changing area does not have to be re-aimed or taken down to stop feeding that corridor into a detection. Both are configured at commissioning and both stay the customer’s call afterwards. What that does not do belongs in the same breath. Masking a zone in the analytics layer does not change what the camera captures or what the recorder keeps, and section 435 is about the recording being made. A mask narrows what gets analyzed and leaves the underlying footage where it was. The audit is still the control. The rest of the limits are the usual ones. Nsightify produces real-time alerts that a condition was observed at a place and a time; it is an alerting layer, not a legal control, and using it resolves no obligation under any statute above. Detection depends on sightlines, mounting height and lighting. Deployment is either Nsightify Cloud or a Zero Trust on-premises installation where video stays inside the site boundary. Nothing here is legal advice. ## Frequently asked questions ### Can my employer put a camera in the break room? No California statute names the break room. Labor Code 435 names three rooms — restroom, locker room, and a room designated for changing clothes — and AB 1331, signed in September 2026, adds a bathroom monitoring prohibition from 1 January 2027. A break room falls outside both, so the question turns on whether employees had a reasonable expectation of privacy there, which is decided on the facts rather than by a list. Confirm your own position with counsel. ### Are cameras allowed in employee-only areas? Generally yes. California law does not restrict cameras by whether an area is open to the public. It restricts them by room: Labor Code 435 prohibits an employer from causing an audio or video recording of an employee in a restroom, locker room or changing room without a court order. A warehouse floor, dock or yard is employee-only and is not covered by that prohibition. ### Is it legal to have cameras in the workplace in California? Yes, with limits. Video surveillance of work areas is lawful in California. Labor Code 435 bars recording an employee in a restroom, locker room or changing room, and bars the employer from using any such recording for any purpose. From 1 January 2027, AB 1331 also bars using a surveillance tool to monitor employees in a workplace bathroom. Audio raises a separate question under Penal Code 632, which requires the consent of all parties to record a confidential communication. ### Does my employer have to tell me about cameras? If your employer is a business covered by the California Consumer Privacy Act, yes, but there is no waiting period. Civil Code 1798.100 requires it to tell you the categories of personal information collected, the purposes, and the retention period at or before the point of collection. No California statute imposes a 14-day or 30-day advance notice before workplace monitoring may begin, despite that claim appearing online. ### Can my boss watch me on camera all day? Continuous monitoring of work areas is not itself prohibited in California. The limits are on place and on use: no recording in the three rooms Labor Code 435 names, no use of such a recording for any purpose, no monitoring in a workplace bathroom from 1 January 2027 under AB 1331, and all-party consent before recording a confidential conversation. What an employer may then do with footage in a disciplinary decision is governed separately. ### Did California pass the bathroom camera bill? Yes. AB 1331 was signed on 30 September 2026 and takes effect on 1 January 2027. It bars an employer from using a workplace surveillance tool, cameras included, to monitor employees in a workplace bathroom, with penalties of up to 500 dollars per violation. It still allows a one-time check of entry and exit, covers no other room, and adds no duty to notify employees. * * * ## A camera map audit to run before AB 1331 takes effect None of this needs to wait for 1 January. Pull the camera list, and for each device write down what is in the frame at the edges, whether it records or only streams, whether audio is enabled, and when the surrounding room was last repurposed. Four columns, one afternoon. Then sort the result into prohibited, defensible with a documented reason, and unexamined — and work the third list first, because that is the one that turns into a complaint nobody can answer. The same inventory does double duty: it is the [stream inventory an IT reviewer will ask for](/nsights/video-surveillance-network-requirements-it-approval/) before any analytics platform is approved, and it is the document that makes the [human-review workflow California’s AI workplace law requires](/nsights/california-ai-workplace-law-sb-947/) possible to write. That law, SB 947, was signed the same day as AB 1331 (Chapter 859, Statutes of 2026). If you want to know what your existing coverage can actually detect once the map is clean, [see what real-time physical safety monitoring reads from the cameras you own](/nsights/physical-safety-monitoring-real-time-ai/). And when the audit leaves you unsure what a particular camera captures, [send the Nsightify team your question](/contact/?interest=question). **Keep reading:** * [Machine guarding restricted areas: where optical detection fits](/nsights/machine-guarding-restricted-zone-optical-detection/) * [OSHA’s warehouse national emphasis program now runs to 2031](/nsights/osha-warehouse-emphasis-program-2026/) More on this from Nsightify: [PPE detection and hazard-zone monitoring](/solutions/physical-safety/). [Previous](/nsights/video-surveillance-network-requirements-it-approval/) [Video Surveillance Network Requirements: Getting IT to Yes](/nsights/video-surveillance-network-requirements-it-approval/) [Next](/nsights/bipa-video-analytics-biometric-identifier/) [BIPA and Video Analytics: Where Biometric Collection Starts](/nsights/bipa-video-analytics-biometric-identifier/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Cargo Theft Statistics 2026 Show Fewer, Bigger Thefts URL: https://nsightify.com/nsights/cargo-theft-statistics-2026/ Cargo theft statistics 2026: in Q2, 26% fewer thefts and losses more than doubled. What that means for loaded trailers in a yard, and what cameras cannot stop. [Home](/) [Nsights](/nsights/) Cargo Theft Statistics 2026 Show Fewer, Bigger Thefts September 25, 2026 11 min. read # Cargo Theft Statistics 2026 Show Fewer, Bigger Thefts AI & Security Perimeter SecurityWarehouse OperationsAI Video Analytics ![A trailer drop lot at dusk, rows of plain parked trailers behind a chain-link fence, with a pole-mounted camera and yard lights and no one in view.](/_astro/cargo-theft-statistics-2026-hero.B8pxRmKG_Z12SePK.webp) Read the cargo theft statistics for 2026 in the wrong order and they look like good news. Verisk CargoNet recorded 26% fewer thefts across the United States and Canada in the second quarter than a year earlier, and estimated losses more than doubled, to $304.6 million. For any site that holds loaded trailers, the question stops being how often thieves come and becomes what is sitting unattended on the night they do. ## Cargo theft statistics for 2026 point two ways at once The headline numbers come from [Verisk CargoNet’s second-quarter release](https://www.globenewswire.com/news-release/2026/08/06/3340253/0/en/cargo-theft-losses-more-than-double-to-304-million-in-q2-despite-a-drop-in-thefts-driven-by-high-value-metals-and-technology-heists.html), published on 6 August. CargoNet recorded 677 theft incidents in the quarter, down 26% from the second quarter of 2025 and 14% from the first quarter of 2026. Estimated losses reached $304.6 million, against $135.7 million a year earlier. Fewer events and larger losses can only mean one thing: on average, each theft is taking more. CargoNet put the average value, among thefts with a reported commodity value, at $564,009. Then it added a caveat: the figure was “heavily influenced by a small number of extreme, multimillion-dollar losses.” That caveat is the story. An average pulled up by a handful of enormous events does not describe a typical theft. It describes a distribution with a long tail, where most incidents are ordinary and a few are catastrophic. The losses are concentrated, not evenly larger. The longer trend runs the same way. [CargoNet estimated 2025 losses](https://www.cargonet.com/news-and-events/cargonet-in-the-media/2025-theft-trends/) at $725 million, up 60% on 2024, with confirmed cargo thefts up 18% and an average value per theft of $273,990. Summarizing 2025, CargoNet’s vice president of operations described organized groups as “targeting extremely high value shipments rather than relying on opportunistic theft.” Cargo theft, as the FBI’s Uniform Crime Reporting Program defines it, is the criminal taking of goods that make up a commercial shipment of freight moving in commerce, from, among other places, a truck, trailer, container, warehouse or distribution facility. The definition treats cargo as moving at every point between origin and destination, including any temporary stop along the way. That last clause is the one that matters for a site. Under the [FBI’s cargo theft definition](https://ucr.fbi.gov/crime-in-the-u.s/2019/crime-in-the-u.s.-2019/additional-data-collections/cargo-theft), a loaded trailer parked in a drop lot overnight is still freight in transit. The yard is part of the route. ## Why incident counts mislead a yard security budget Site security budgets are usually argued from frequency. How many incidents did we have last year? How many did the region have? Is the trend up or down? That logic worked when losses were many and small, because the incident count tracked the cost. It stops working once a few events carry most of the loss. **The count invites deferral.** A falling incident rate reads as a reason to wait a year. The same quarter that produced 26% fewer thefts produced more than double the losses. A security lead who reports only the first number to finance is reporting good news that is not true. **The average invites the wrong budget.** Divide a loss total by an incident count and you get a figure that describes no actual event. Plan around $564,009 per theft and you overestimate most incidents and badly underestimate the one that matters. **One control is asked to stop three different crimes.** A fence, a camera and a guard address a thief who physically takes a trailer. They do nothing about a thief who arrives with convincing paperwork and is handed the load at the dock. They do nothing about one who redirects a shipment by compromising a carrier’s email. Treating cargo theft as one problem means buying one kind of answer and assuming it covers the rest. ## Plan yard security around the high-value load The practical shift is small to describe and harder to do. Stop asking how often theft happens and start asking which loads on your site would be a disaster to lose, where they sit, and for how long. ### Split the count into the crimes it contains CargoNet’s release separates the categories, and they are moving differently. The split decides which controls are even relevant. Crime What changed in Q2 2026 What actually reaches it Events CargoNet classifies as theft 378 events, down from 488; CargoNet saw fewer organized thefts of unattended trailers Locks, yard layout, after-hours detection and a response plan Fictitious pickup: a thief poses as the carrier 158 events, against 165 a year earlier Carrier and broker vetting, driver identity checks, secure pickup numbers Email compromise and shipment misdirection Persistent, per CargoNet; comparatively stable in California and Texas; no count published Cybersecurity and payment and dispatch verification Two readings follow. Events CargoNet classifies as theft are falling but are still the largest category, 378 of 677. Fictitious pickups barely moved, and CargoNet calls account compromise and misdirection persistent. Part of the fall came from fewer non-delivery schemes, which no camera sees either. Any honest yard security plan has to carry both facts at once. ### Price the tail, not the average If the loss is concentrated in a few events, the planning unit is the high-value load, not the typical incident. CargoNet’s commodity data shows where thieves are aiming. Metal thefts rose from 54 in the second quarter of 2025 to 80 in the second quarter of 2026, with copper the most targeted. Enterprise technology continued to draw organized groups. Beverages and grocery fell by 36 events combined; seafood rose by 11. The exercise for a site is short. List the loads that pass through in a typical week and mark the handful whose loss would be a board-level event, not an insurance claim. Then note how long each one sits in the yard between arrival and departure. The tail risk lives in that short list, and it moves week to week with what you are holding. > The unit of exposure is not the incident. It is the loaded trailer nobody is watching. ### Treat the unattended loaded trailer as the thing to protect [Travelers’ guidance on cargo theft tactics](https://www.travelers.com/resources/business-topics/supply-chain-management/5-tactics-cargo-thieves-use) starts in the same place: avoid unattended loaded trailers where possible, especially in high-theft areas. The insurer also recommends high-security rear door locks and air cuff locks, and landing gear locks where staging or dropping a load cannot be avoided. Each one turns a quick hitch-and-drive into something that takes time and tools. Those controls delay a theft. They do not tell anyone it is happening. A lock that holds for several minutes is only worth those minutes if someone learns within them that a vehicle is backed up to the trailer. Delay and detection work as a pair, and a yard that has one without the other has bought half a control. ### Watch the calendar, including the days you are open CargoNet’s [Labor Day analysis, published on 4 September](https://www.globenewswire.com/news-release/2026/09/04/3356766/0/en/verisk-cargonet-reports-persistent-labor-day-cargo-theft-exposure-across-u-s-freight-networks.html), covered five years, 2021 to 2025. It counted each holiday window from the Thursday before Labor Day to the Wednesday after. Across 273 thefts, the four weekdays in that window accounted for 194, or 71%. Saturday through Monday accounted for 79. Per day, that is an average of about 48.5 thefts per weekday against about 26 per weekend day, over the five years combined. The busiest days were not the ones when the yard was emptiest. The annual count also climbed, from 33 in 2021 to 56 in 2025, with a peak of 70 in 2024. California, Texas and Illinois together accounted for 130 of the 273. CargoNet’s own reading is that the weekday pattern fits the growing share of theft carried out through deceptive pickups and non-delivery schemes, which need working offices and phone lines to succeed. It also names the physical exposure of freight left sitting while sites are closed. So two questions are worth asking about your own site: who confirms carriers and pickups on the days around a holiday, and which loads are staged before a long weekend and collected after it? Thanksgiving falls on 26 November this year, and the same question applies to the whole year-end period. ### Three yard conditions worth an after-hours alert A yard does not need to watch everything. It needs to know quickly about the few conditions that precede a physical theft. 1. **A vehicle stopped where nothing should stop.** Along the fence line, at a closed gate, beside the drop lot, outside operating hours. 2. **A person inside the yard boundary after hours**, or lingering along the fence where the loaded trailers are parked. 3. **Something leaving that was supposed to stay.** Goods or equipment removed from where they were staged, or a vehicle heading out through the gate outside operating hours. Each of these is a visual fact in a scene. Each one is also useless as an alert unless it reaches a person who can act, which is why [calibrating loitering detection to cut false alarms](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/) matters as much as detecting anything at all. An operator who clears false alerts all night learns to treat every one as nothing, the pattern known as [alert fatigue in perimeter security](/nsights/decision-models-alert-fatigue-perimeter-security/). ## What after-hours yard detection can and cannot do Nsightify runs on the yard’s existing IP and CCTV cameras. The detections that map to the conditions above are part of [perimeter security monitoring on existing cameras](/solutions/perimeter-security/): after-hours zone entry, loitering, stopped vehicle, restricted-area dwell, directional tripwire crossing, and person and vehicle classification, all active from day one. Removed object and perimeter pacing are included and switched on at commissioning, once they are tuned against the site’s own footage. Each alert reaches the people responsible in real time, which is the half of the delay-and-detection pair that locks cannot provide. The limits are specific. Nsightify does nothing about fictitious pickups: it does not check a driver’s paperwork, read a bill of lading or confirm a pickup number. It does nothing about email compromise or shipment misdirection. On the physical side, detection depends on camera placement, lighting and weather, and lighting that is adequate for people crossing the yard is not always adequate for detection at the far fence. Detection is also not prevention. An alert buys time only if someone acts on it. The way to find out what your own cameras can do after dark is to [test the analytics on your own cameras](/nsights/how-accurate-is-ai-video-analytics-testing-standard/) before relying on them. Whether the light reaches the far fence can be checked before that test, and [video analytics at night](/nsights/video-analytics-at-night/) shows how. ## Frequently asked questions ### Is cargo theft going up or down? Both, depending on the number. Verisk CargoNet recorded 677 theft incidents in the United States and Canada in the second quarter of 2026, 26% fewer than a year earlier. Estimated losses for the same quarter were $304.6 million, against $135.7 million in the second quarter of 2025. For 2025 as a whole, CargoNet estimated losses of $725 million, up 60% on 2024. ### What is a fictitious pickup? A fictitious pickup is a theft in which someone poses as a legitimate carrier and is handed the load. Nothing is broken into. Verisk CargoNet counted 158 fictitious pickups in the second quarter of 2026, against 165 a year earlier. Travelers recommends vetting carriers and brokers and, at pickup, confirming driver identity, truck and trailer details and a secure pickup number. ### When does cargo theft happen around holidays? Often on the working days around the holiday, not the holiday itself. Verisk CargoNet analyzed Labor Day windows from 2021 to 2025, counting from the Thursday before to the Wednesday after. The four weekdays in that window accounted for 194 of 273 thefts, or 71%. Saturday through Monday accounted for 79. CargoNet links the pattern to deceptive pickup and non-delivery schemes, which depend on working offices. ### Where does most cargo theft happen? The 2026 releases from Verisk CargoNet do not publish a breakdown by location type. Its Labor Day analysis describes theft clustering around dense freight networks, large consumer markets, intermodal infrastructure and places where stolen goods can be resold quickly. Percentages circulating online that assign a share of theft to warehouses or parking lots could not be traced to a dated primary source. ### How do you secure a truck yard at night? Start by reducing what is exposed. Travelers recommends avoiding unattended loaded trailers where possible, fitting high-security rear door locks and air cuff locks, and adding landing gear locks when staging loads is unavoidable. Then make sure someone learns quickly when a vehicle or person is in the yard after hours, and has a plan to respond before the load is gone. * * * ## What to check before Thanksgiving Pull last quarter’s inbound and outbound schedule and mark every high-value load that sat in the yard overnight, and for how long. Then check the cameras covering those spots after dark, not at noon. Most sites can do that with the cameras already on the poles, which is the point of [adding AI video analytics to cameras you already own](/nsights/you-dont-need-new-cameras-ai-video-analytics/) rather than starting over. Whatever that night check leaves unanswered, [ask the Nsightify team about your yard](/contact/?interest=question). More on this from Nsightify: [AI video analytics on existing IP and CCTV cameras](/solutions/). [Previous](/nsights/ppe-detection-what-cameras-cannot-verify/) [PPE Detection: What a Camera Verifies, and What It Cannot](/nsights/ppe-detection-what-cameras-cannot-verify/) [Next](/nsights/employee-monitoring-laws-by-state-cameras/) [Do Employee Monitoring Laws by State Cover Cameras?](/nsights/employee-monitoring-laws-by-state-cameras/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Could Decision Models Fix Alert Fatigue in Perimeter Security? URL: https://nsightify.com/nsights/decision-models-alert-fatigue-perimeter-security/ A new AI model makes snap yes-or-no calls on text. Could decision models rank perimeter alerts by context and ease alert fatigue? The case on both sides. [Home](/) [Nsights](/nsights/) Could Decision Models Fix Alert Fatigue in Perimeter Security? By Ricardo Assaad October 7, 2026 14 min. read # Could Decision Models Fix Alert Fatigue in Perimeter Security? AI & Security Perimeter SecurityAI Video AnalyticsComputer Vision ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_ieouy.webp) A fence-line camera flags a person at 3 a.m. The operator has cleared two hundred alerts tonight, and every one was nothing. This one gets the same two seconds as the rest. That moment is where a lot of perimeter security succeeds or fails, and a sharper camera doesn’t change it. In September a new kind of AI model arrived that is built for this sort of snap judgment. Nobody has shown that it works for security. This piece sets out the argument on both sides, and the questions I can’t answer yet. ## False alarms are older than the cameras False alarms were swamping security long before anyone put AI on a camera. A [U.S. Justice Department guide on false burglar alarms](https://portal.cops.usdoj.gov/resourcecenter/content.ashx/cops-p014-pub.pdf), first published in 2007 and updated in 2011, reports that between 94 and 98 percent of alarm calls were false. Those calls made up 10 to 25 percent of all calls to police. In 2002 alone, U.S. police answered about 36 million alarm activations at an estimated cost of $1.8 billion. Hospitals have the same problem and gave it a name. [The Joint Commission’s 2013 alert on medical device alarms](https://digitalassets.jointcommission.org/api/public/content/f65e5c9df2b94000a99445e0a7877007) cited estimates that 85 to 99 percent of alarm signals need no clinical intervention. It called the result alarm fatigue, and it counted 98 reported alarm-related events over three and a half years, 80 of them fatal. The pattern is the same in both fields. When nearly every alert is nothing, people learn to treat every alert as nothing, and the real one arrives looking like all the others. Those figures are old and come from neighboring fields. I have not found an independent, published false-alert rate for AI video analytics on perimeters. Most of the numbers in circulation come from vendors selling the fix, and that missing baseline is the first open question in this piece. ## Why better detection hasn’t ended it Two things keep the alert count high even as detectors improve. The first is arithmetic. Picture a site that logs 10,000 motion events a week, and one of them is a real intrusion. Give that site a detector that is right 99% of the time in both directions. It still raises about 100 false alerts for the one real one, so roughly 99 in every 100 alerts are noise. Those numbers are an illustration, not a measurement, but rare events do this to any detector, however good. The second is that the detector answers a narrower question than the operator’s. It can say “person, north fence, 3:12 a.m.” It can’t say whether that person is the night guard on his usual round, a contractor who stayed late, or someone who just cut the fence. That depends on things outside the frame: the shift roster, the access log, the contractor schedule, whether gate 2 was left open. Today the gap between detecting and deciding is filled by fixed rules and by the operator’s attention. Rules can’t anticipate every exception, and attention wears down over a shift. ## What a decision model is On September 15, a startup called TypeSafe AI opened early access to Jev, a model that reads text the way a chatbot does and cannot write a word. You hand it a block of context and a set of typed questions. It returns a yes/no probability, a score on a scale you define, or a pick from options you list, each with a probability attached. [TypeSafe’s launch post](https://typesafe.ai/blog/introducing-system-one-models-and-jev) says it answers in 70 to 500 milliseconds and charges $0.042 per million input tokens, with output free. It also says the probabilities are calibrated, meaning that when the model says 80%, it should be right about 80% of the time. No paper or model weights back that claim yet. _Current as of October 7, 2026: Jev is at version 1.13. It takes text only, runs only on TypeSafe’s hosted service, and has no published paper or weights. TypeSafe says later versions will fix some of the weaknesses described below._ Jev can’t see. It takes text only, so in a security system it would sit after the detector and read a description of each event. Something like this: * **Event:** person, north fence zone, 03:12, walking along the fence line for 40 seconds. No badge-in during the last hour. Gate 2 closed. Contractor window ended at 18:00. * **Question 1:** Is this person likely to be authorized? (yes/no) * **Question 2:** How urgent is this, from 0 to 5? * **Question 3:** Who should get it: the guard desk, the supervisor, or the log? TypeSafe calls these System One models. Commentators, Simon Willison among them, prefer “decision models”, and the idea has older relatives. Classifiers and rerankers have made this kind of call for years. What’s new is asking arbitrary questions in plain language with no training step. ## The case that it helps **It aims at the right gap.** The context an operator is missing mostly exists as text already: rosters, access logs, contractor schedules, site rules. A model that reads text and returns a judgment fits the hole between the detector and the person. **Probabilities let you rank instead of filter.** A rule either fires or it doesn’t. A probability lets you sort the queue so the alert most likely to matter sits on top, and nothing has to be hidden. The 3 a.m. operator checks the top of a list instead of clearing all of it. **The price makes it possible to judge every event.** A site sending 10,000 events a day, at roughly 1,000 tokens each, would spend about 42 cents a day at TypeSafe’s list price. That is my arithmetic on their published rate, with assumed volumes. At that cost you don’t have to sample. **Site rules could be written in plain language.** Every site has its own exceptions. A safety manager could write “nobody on the loading dock roof without a second person” and have it applied the same day, without waiting for a vendor to code it. ## The case that it doesn’t **It may be aimed at the wrong errors.** If most false alerts come from perception (a shadow, a deer, [headlights sweeping a fence at night](/nsights/video-analytics-at-night/)), a text model downstream can’t help. It sees only what the detector told it, so a mislabeled event gets judged as labeled. **Security events are made of numbers.** TypeSafe’s own documentation says the current version struggles with numbers, dates and double negatives. Timestamps, dwell times, counts and distances are most of what an event record contains. **A wrong call costs far more in one direction.** Ranking a real intrusion at the bottom of the queue is a miss, and nobody finds out until later. A triage model that is wrong one time in twenty may suit support tickets and still be unacceptable on a fence line. It also always answers: if none of the options you listed fits, it picks one anyway. **It gives a number and no reason.** After an incident, “the model scored it 0.12” won’t satisfy an investigator or an insurer. In [his write-up of Jev](https://simonwillison.net/2026/Sep/21/jev/), Simon Willison had the model answer “Good city?” for each Bay Area city, and it put Cupertino on top and East Palo Alto at the bottom. He used it to warn that a bare score can hide bias nobody sees, and concluded that testing matters even more for a model like this than for ordinary large language model (LLM) projects. The security version of that risk is a score that quietly depends on how a person was described. **It can be talked into an answer.** One review of TypeSafe’s stated limitations, from Respan, a company that sells a competing classifier, says the model doesn’t treat its input as hostile by default, and that injected text can move the result. Most software meets that risk now and then. A security system has an adversary by design, and any text an outsider can influence, such as a visitor log entry, is a possible lever. **It is closed, remote and three weeks old.** Jev is a hosted service from one vendor. Many sites can’t send event data off the premises, and every site needs alerts to keep working when the link drops. The same review notes that model updates can shift the scores a threshold was tuned on. Open-source alternatives appeared within days, though early reports put most of them behind on quality. ## What it would have to beat A decision model has competition, most of it older and some of it already in the field. Approach Reads Where it’s strong Where it’s weak Fixed rules (zones, schedules, [dwell thresholds](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/)) Event fields Predictable and explainable, with no running cost Brittle. Someone has to write each exception for each site Classifier trained on a site’s own alert history Event fields Accurate once trained, and runs on site Needs labeled history, which a new site doesn’t have Vision-language model reviewing the clip Pixels Can catch what the detector described wrongly Slower and costlier per event General-purpose LLM reading the event Text Works through several steps and explains itself Seconds per call, costly at volume, shaky confidence scores Decision model Text Fast and cheap, with a probability on every answer No reasons, weak on numbers, untested in this field The vision-language row deserves attention. At an industry conference in June, an integrator described perimeter systems at high-security sites, including U.S. government facilities, that combine signal analytics, computer vision and vision-language models. [Its session abstract](https://prevent.securityindustry.org/events/transforming-perimeter-security-operations/) promised a dramatic cut in alarms, with no baseline or measurement period behind it, and I can’t check it. If results like that hold, the strongest rival is a model that looks at the footage a second time. The comparison I’d most like to see is the least glamorous one: a decision model against a well-kept rules engine plus a small classifier trained on the site’s own history. ## What a fair test would look like Any site with a labeled alert history can test this, much as it would [test a detector’s accuracy on its own cameras](/nsights/how-accurate-is-ai-video-analytics-testing-standard/), and the test is small. 1. **Collect a few hundred logged events,** each with an operator’s verdict: real, benign or unclear. 2. **Build the baseline first.** Write the rules you would actually deploy and record how they sort the same events. 3. **Describe each event as text,** ask the model the triage questions, and rank the events by its probabilities. 4. **Count how many alerts an operator must open,** working down each ranking, to reach every real event. Note whether any real event lands near the bottom. 5. **Check calibration.** Of the events the model scored at 90%, about nine in ten should have been right. 6. **Try to break it.** Change timestamps and counts, and plant instructions in the free-text fields. A model that wins step 4 and survives step 6 has earned a pilot. One that only ties the rules adds cost and risk for nothing. I haven’t run this test yet. ## Questions I can’t answer yet * **How many false alerts are perception errors, and how many are context errors?** If a site’s noise is mostly shadows and animals, no text model will touch it. I don’t know the split, and I doubt any two sites share one. * **Who is accountable when triage is wrong?** If a ranking layer puts a real intrusion last and it gets missed, the fault could sit with the integrator, the software vendor, the model provider or the site. I don’t know whether today’s contracts settle that. * **Does a ranked queue help operators, or teach them to skip?** Sorting should focus attention on the top of the list. It might also give people permission to ignore the bottom half, where a missed intrusion would be hiding. * **Does calibration survive a new domain?** A probability that is honest on support tickets may be less honest on fence-line events the model never saw in training. * **Should a safety manager write detection policy in plain English?** It removes a bottleneck. It also removes the engineer who would have asked what the rule does at shift change. * **Would an insurer or an auditor accept a probability as a record?** A rule can be read back after an incident. A score can only be reported. * **Does cheaper judgment mean fewer alerts, or more?** TypeSafe named its model after William Stanley Jevons, the economist who saw that more efficient steam engines raised the demand for coal. Cheap decisions could tempt sites to add cameras and rules until the queue is as long as it was. ## Where I land for now I build video analytics at Nsightify, so I have a stake in this gap, and you should weigh my view with that in mind. My working guess is that a decision model could earn a place as a ranking layer for alerts that turn on context. I expect it to do nothing for perception errors, and I wouldn’t yet let one hide an alert from a person. All three are guesses until someone runs the test. I’m more confident about where the remaining work is. Detection has improved faster than the judgment that follows it, and the 3 a.m. operator is still making that judgment alone. If you run a site with a labeled alert history and want to see how the test comes out, I’d like to compare notes. [Reach me through the contact form](/contact/?interest=question). * * * ## Frequently asked questions ### What is alert fatigue? Alert fatigue, which hospitals call alarm fatigue, is what happens when nearly every alert turns out to be nothing: people learn to treat every alert as nothing, and the real one arrives looking like all the others. The Joint Commission’s 2013 alert on medical device alarms cited estimates that 85 to 99 percent of alarm signals need no clinical intervention. A U.S. Justice Department guide reports that 94 to 98 percent of burglar alarm calls were false. ### What is a decision model in AI? A decision model reads text and returns a judgment instead of writing anything back: a yes/no probability, a score on a scale you define, or a pick from options you list, each with a probability attached. TypeSafe AI released one, Jev, in early access on September 15, 2026, and calls it a System One model. Classifiers and rerankers have made this kind of call for years. What’s new is asking arbitrary questions in plain language with no training step. ### Why do accurate detectors still produce mostly false alerts? Because real events are rare. Picture a site that logs 10,000 motion events a week, one of them a real intrusion, watched by a detector that is right 99% of the time in both directions. It still raises about 100 false alerts for the one real one, so roughly 99 in every 100 alerts are noise. The numbers are an illustration, not a measurement, but rare events do this to any detector. ### Can a decision model reduce false alarms from perimeter cameras? Only the ones that turn on context. A decision model reads a text description of each event, so it can weigh a shift roster, an access log or a contractor schedule, but it cannot see the footage. A false alert caused by a shadow, an animal or headlights gets judged exactly as the detector labeled it. Nobody has yet shown that decision models work for security alerts. ### How would a site test a decision model on its own alerts? Start with a few hundred logged events that operators have already marked real, benign or unclear. Write the rules you would actually deploy and record how they sort the same events. Then describe each event as text, rank the events by the model’s probabilities, and count how many alerts an operator must open to reach every real one. Check calibration, then try to break it by changing timestamps and planting instructions in free-text fields. ## Sources * TypeSafe AI, [Introducing System One Models & Jev](https://typesafe.ai/blog/introducing-system-one-models-and-jev), September 15, 2026 * TypeSafe AI, [Jev 1.13 jaggedness](https://docs.typesafe.ai/model-jaggedness/jev-1.13), product documentation, last reviewed October 2, 2026 * Zapier, [What is Jev? TypeSafe AI’s System One model](https://zapier.com/blog/jev/), September 24, 2026 * Sébastien Dubois, [Jev and System One models - a semantic if statement for your code](https://www.dsebastien.net/jev-and-system-one-models-a-semantic-if-statement-for-your-code/), September 23, 2026, updated October 4, 2026 * Respan, [What Is the Jev AI Model? Pricing & Use Cases (2026)](https://www.respan.ai/articles/what-is-the-jev-ai-model), September 23, 2026 * Simon Willison, [Jev introduces a new shape of LLM—System One, aka Decision Models](https://simonwillison.net/2026/Sep/21/jev/), September 21, 2026 * Rana Sampson, [False Burglar Alarms, 2nd Edition](https://portal.cops.usdoj.gov/resourcecenter/content.ashx/cops-p014-pub.pdf), Problem-Specific Guides Series No. 5, U.S. Department of Justice, Office of Community Oriented Policing Services, March 2007, updated August 2011 * The Joint Commission, [Sentinel Event Alert, Issue 50: Medical device alarm safety in hospitals](https://digitalassets.jointcommission.org/api/public/content/f65e5c9df2b94000a99445e0a7877007), April 8, 2013 * Security Industry Association, [From Noise to Intelligence: Transforming Perimeter Security Operations](https://prevent.securityindustry.org/events/transforming-perimeter-security-operations/), session abstract, Perimeter PREVENT 2026, June 24, 2026 Ricardo Assaad is the founder and CEO of Nsightify. More on this from Nsightify: [AI video analytics on existing IP and CCTV cameras](/solutions/). [Previous](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) ## More on topic [All Nsights](/nsights/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) [ ![A dome security camera mounted above an empty warehouse loading dock with the dock doors closed and a painted floor line.](/_astro/employee-monitoring-laws-by-state-cameras-hero.DTXNao5P_Z27osnE.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ Physical Safety Regulatory Compliance ](/nsights/employee-monitoring-laws-by-state-cameras/) [Do Employee Monitoring Laws by State Cover Cameras?](/nsights/employee-monitoring-laws-by-state-cameras/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Do Employee Monitoring Laws by State Cover Cameras? URL: https://nsightify.com/nsights/employee-monitoring-laws-by-state-cameras/ Employee monitoring laws by state treat cameras by purpose. What Connecticut requires from 1 October, how Maine leaves safety cameras out, and what to write. [Home](/) [Nsights](/nsights/) Do Employee Monitoring Laws by State Cover Cameras? September 28, 2026 13 min. read # Do Employee Monitoring Laws by State Cover Cameras? Physical Safety Regulatory CompliancePerimeter Security ![A dome security camera mounted above an empty warehouse loading dock with the dock doors closed and a painted floor line.](/_astro/employee-monitoring-laws-by-state-cameras-hero.DTXNao5P_Z1Cjy4n.webp) From 1 October, a Connecticut employer running cameras on its own floor must tell employees where it monitors, not just what kind of monitoring it does, unless a security and safety exception applies. In Maine, a newer law leaves a camera used for security or safety out of scope entirely. Employee monitoring laws by state disagree about cameras, and the thing they disagree on is what the camera is for. ## Employee monitoring laws by state were written for inboxes Search for employee monitoring laws by state and the first page is a set of fifty-state tables. They list Connecticut, New York and Delaware as notice states, add Maine for 2026, and treat every entry as the same kind of rule. Several are published by companies that sell keystroke and screen tracking software, and the tables are built for that product. A camera is a different question. Whether a state’s notice law reaches a camera at all is decided in the statute’s definitions, and the definitions do not agree. Some name cameras and then carve out a purpose. One names photo-optical systems only as a way of intercepting email and phone calls. One excludes security and safety cameras from its definition altogether. An employee monitoring notice law requires an employer to tell employees what electronic monitoring it conducts, before or when the monitoring begins. The state laws differ on what counts as monitoring, what the notice must say and which purposes are exempt, and those differences decide whether a workplace camera is covered at all. For a company running cameras in more than one state, that makes the camera’s purpose the fact everything else depends on. It is also the fact least often written down. _Status, current as of 24 September 2026: Connecticut’s rewrite takes effect on 1 October, with no delay or amendment announced. Nothing below is legal advice, and each statute is linked so you can read it yourself._ ## Three shortcuts that trip up multi-state camera programs Three habits get multi-state camera programs into trouble, and each one is a reasonable shortcut. **The fifty-state table.** It answers “is there a notice law?” and stops. It does not say whether the law reaches cameras, and it rarely separates a camera on a loading dock from software on a laptop. One 2026 guide that ranks for these searches lists Maine’s law as effective in March. [The Maine Department of Labor](https://www.maine.gov/labor/news_events/article.shtml?id=13356290) puts the date at 29 July. Copy the table into a compliance calendar and the calendar inherits the error. **One sign at the front door.** Plenty of sites post a single notice at the entrance saying the premises are monitored, and treat it as covering everything. From 1 October that stops matching Connecticut’s text, which wants the specific locations named and the notice posted at the location as well. **Treating every camera the same.** A lobby camera, a dock camera and a camera on a production line can sit on the same recorder and fall under different rules in the same state. The difference is not the hardware. It is what each camera is for and where it points, and a program that has never written that down has nothing to test against a statute. ## Reading each statute by what the camera is for The useful exercise is to take the four statutes that come up most often and ask the same three questions of each. Does it reach a camera? What must the notice say? And is there a purpose-based exception? State What the text says about cameras What the notice must cover Purpose-based exception Connecticut, from 1 Oct 2026 Names cameras; excludes collection for security purposes in common areas open to the public Types of monitoring and specific locations; posted, including at the location Location disclosure waived for security and employee safety purposes Maine, since 29 Jul 2026 Excludes cameras used for security or safety Notice before surveillance, at the hiring interview and once a year Security and safety cameras sit outside the definition New York Names photo-optical systems only as a means of monitoring phone, email or internet use Written notice on hiring, acknowledged and posted A camera watching a work area is not a listed subject Delaware Telephone, email and internet only; video not mentioned Daily electronic notice, or a one-time acknowledged notice Video is not mentioned California sits outside this pattern. For businesses the CCPA covers, the notice duty for workplace cameras runs through the CCPA’s notice at collection rather than a monitoring statute, and [where California’s workplace surveillance laws say cameras cannot go](/nsights/california-workplace-surveillance-camera-placement/) covers the state in more detail. ### Connecticut from 1 October: types, locations, and a sign where it happens Connecticut has required notice of electronic monitoring since 1998, and its definition has always named cameras and photo-optical systems. [Public Act 26-73](https://www.cga.ct.gov/2026/act/Pa/pdf/2026PA-00073-R00SB-00472-PA.PDF), approved on 4 June 2026, rewrites the statute from 1 October. Three duties are added, plus a new exception. 1. **Location, not just type.** Prior written notice must now name the types of monitoring and the specific locations on the premises where it may occur. 2. **A notice at the location.** The posted notice must go in a conspicuous place, including in the specific location where monitoring may occur. The posting still counts as the written notice. 3. **A statement for new hires.** Employees hired on or after 1 October must receive, before they start, a plain-language statement of the activities that are prohibited and may be monitored without notice. The penalties are unchanged: up to $500 for a first offense, $1,000 for a second and $3,000 for each one after, levied by the Labor Commissioner after a hearing. The long-standing exception also survives. An employer with reasonable grounds to believe employees are breaking the law, violating the legal rights of the employer or its employees, or creating a hostile workplace may monitor without notice where monitoring may produce evidence of it. ### The two Connecticut exceptions that decide most camera questions The first exception is in the definition. Collection of information for security purposes in common areas of the premises held out for use by the public is not electronic monitoring at all. The exclusion’s words reach a security camera in a lobby held out for public use. They do not reach a camera over a dock the public does not use. The second is new. The requirement to disclose the specific location does not apply at an airport, or when the employer has reasonable grounds to monitor “for security and employee safety purposes.” Two details in that sentence matter. It waives the **location**, not the notice. The duty to disclose the types of monitoring still applies, so a safety camera exempt from the location rule still belongs in the notice. And it says **and**. The waiver reads as covering monitoring for security and employee safety, not security or safety. Whether a camera used only for safety, with no security purpose, qualifies is a question for counsel. So is whether the waiver also removes the duty to post at the location, since that duty is also a location requirement. The statute does not answer either on its face. ### Maine: a notice law that leaves safety cameras out [Maine’s employer surveillance law](https://legislature.maine.gov/legis/bills/getPDF.asp?paper=HP0025&item=3&snum=132), Public Law 2025, chapter 524, became law without the Governor’s signature on 11 January 2026. The Maine Department of Labor gives its effective date as 29 July 2026. It covers monitoring through photo-optical systems, and then excludes an employer’s use of “surveillance cameras for security or safety purposes.” Where the law does apply, the employer must notify employees before surveillance begins, tell candidates during the hiring interview and give written notice once a year. It also restricts audiovisual monitoring in an employee’s home, personal vehicle or property unless the job requires it. Fines run from $100 to $500 per violation. Set Maine beside Connecticut and the conjunction changes. Maine says security **or** safety. Connecticut’s waiver says security **and** employee safety. Maine’s exclusion names safety on its own, so its words reach a camera installed only to watch for blocked exits. Connecticut’s text is less clear, which is why the purpose you write down matters more there, not less. ### New York and Delaware: phone, email and internet, not cameras [New York Civil Rights Law section 52-c](https://www.nysenate.gov/legislation/laws/CVR/52-C*2) is often listed alongside Connecticut and Delaware as a monitoring notice law. Its notice duty attaches when an employer monitors or intercepts telephone conversations, email, or internet access and usage. It names photo-optical systems, but only as one of the means by which that monitoring might be done. A camera watching a work area is not monitoring any of the three. The notice must be given on hiring, acknowledged in writing or electronically, and posted, with penalties of up to $500, $1,000 and $3,000. [Delaware’s statute, title 19, section 705](https://delcode.delaware.gov/title19/c007/sc01/index.html#705), is narrower still. It covers telephone, email and internet usage, offers a choice between a daily electronic notice and a one-time acknowledged notice, and does not mention video. > The hardware is the same in every state. The notice duty follows what the camera is for. ### When a safety camera starts answering other questions Every purpose-based line above depends on the purpose staying put. It often does not. A camera installed to watch for blocked exits gets asked, months later, who was on the dock at 3 a.m. Footage pulled for a safety review ends up in a disciplinary file. A count of people at a door, useful for crowd limits, becomes a throughput number on a manager’s report. None of the four statutes above prohibits these uses, and none of them says a change of use removes an exception. But Maine and Connecticut both draw their line by purpose, and a camera whose actual use has drifted away from its written purpose is harder to defend on that line. If camera alerts start feeding decisions about individual workers, the question also moves into [the human-review step California’s SB 947 would require](/nsights/california-ai-workplace-law-sb-947/) (on the Governor’s desk as of 24 September 2026). That is a different body of law, answering a different question. ### Writing the purpose statement once for every site The document that holds a multi-state program together is short. For each group of cameras with the same job, write down: 1. **Where the cameras are**, by area, not just by site. 2. **What they are for**: security, employee safety, both, or something else, in plain words. 3. **Whether the area is open to the public**, which in Connecticut can take a security camera out of scope. 4. **What analytics run on them**, if any, and what each one reports. 5. **Who receives alerts and footage**, and whether either is used for discipline or performance. 6. **When the statement is reviewed**: at least annually, and whenever a camera’s use changes. The same page then answers each state’s question in its own terms. Connecticut asks for types and locations. Maine asks whether the camera is for security or safety. New York and Delaware ask whether phone, email or internet use is monitored. If a camera group also runs features that measure faces, that is a separate notice duty altogether, covered in [where biometric collection starts under BIPA](/nsights/bipa-video-analytics-biometric-identifier/). ## What camera analytics changes about the notice question Nsightify adds real-time analytics to a site’s existing IP and CCTV cameras. It does not add a camera, so the notice question attaches to the existing estate. What analytics can change is what a camera is used for, which is why the purpose statement should name every detection running on each group. Nsightify can also exclude a camera from analysis, or mask a zone within a view so it is never analyzed, so that what runs on each group matches the purpose written down. That limits analysis, not recording: masking changes neither what the camera captures nor what the recorder keeps. Most of Nsightify’s detections describe a security or safety condition in a scene: PPE compliance, a blocked exit or obstructed egress path, forklift–pedestrian proximity, a person down, intrusion into a restricted zone, after-hours zone entry. People and vehicle counting is the one to think about carefully. A count at a dock door is a security measure in one purpose statement and a productivity measure in another, and the statutes above care which one it is. The limits are plain. Nsightify is an alerting layer. It does not write notices, post signs or decide which statute applies, and running it resolves no notice obligation. Detection also depends on camera placement, lighting and sightlines. Whether a given deployment falls inside Connecticut’s waiver or Maine’s exclusion is a question for the employer and counsel, and this article does not answer it. ## Frequently asked questions ### Do employers have to notify employees of security cameras? It depends on the state and on what the camera is for. Connecticut’s text names cameras but excludes collection for security purposes in common areas open to the public. Maine’s excludes cameras used for security or safety. New York’s and Delaware’s notice laws are written around telephone, email and internet monitoring. In California, for businesses the CCPA covers, the duty comes through its notice at collection. Confirm your own position with counsel. ### Does Connecticut’s electronic monitoring law apply to cameras? Its definition names them. Connecticut defines electronic monitoring to include cameras and photo-optical systems, and has since 1998. It excludes collection for security purposes in common areas held out for public use. From 1 October 2026 the notice must name the types of monitoring and the specific locations where it may occur, though the location disclosure does not apply where the employer has reasonable grounds to monitor for security and employee safety purposes. ### What changes in Connecticut on October 1, 2026? Public Act 26-73 rewrites Connecticut’s monitoring statute. Employers must give prior written notice of the types of monitoring and the specific locations where it may occur, and post the notice including at the location itself. Employees hired on or after 1 October must receive, before they start, a written plain-language statement of which activities are prohibited and may be monitored without notice. Penalties remain up to $500, $1,000 and $3,000 for first, second and later offenses. ### Does Maine’s employer surveillance law cover security cameras? Its definition excludes them when they are used for security or safety. Maine’s law, in effect since 29 July 2026, defines employer surveillance to include photo-optical systems and then excludes an employer’s use of surveillance cameras for security or safety purposes. Other surveillance requires notice before it begins, disclosure during the hiring interview and written notice once a year, with fines of $100 to $500 per violation. ### Does New York’s electronic monitoring law cover video cameras? Its notice duty is written around telephone, email and internet monitoring. New York Civil Rights Law section 52-c requires notice when an employer monitors or intercepts telephone conversations, email or internet access and usage. It names photo-optical systems only as one means of doing that. The notice must be given on hiring, acknowledged by the employee and posted, with penalties of up to $500, $1,000 and $3,000. * * * ## What to do before 1 October If you have a site in Connecticut, walk it with the camera list in hand and mark each camera by where it is and what it is for. The statute’s text excludes collection for security purposes in common areas held out for public use. It waives the location disclosure where the employer has reasonable grounds to monitor for security and employee safety. For the rest, it asks for the location in the notice and a posting where the monitoring happens. Take the marked-up list to counsel. Then write the one-page purpose statement for every site, so the next state that passes a camera notice law is a question you can already answer. If a camera on that list leaves you unsure what it records or reports, [send that question to the Nsightify team](/contact/?interest=question). More on this from Nsightify: [PPE detection and hazard-zone monitoring](/solutions/physical-safety/). [Previous](/nsights/cargo-theft-statistics-2026/) [Cargo Theft Statistics 2026 Show Fewer, Bigger Thefts](/nsights/cargo-theft-statistics-2026/) [Next](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## How Accurate Is AI Video Analytics? Test It, Don't Trust It URL: https://nsightify.com/nsights/how-accurate-is-ai-video-analytics-testing-standard/ How accurate is AI video analytics? A single percentage cannot say. What IEC 62676-6 and UK government testing measure instead, and how to test a vendor. [Home](/) [Nsights](/nsights/) How Accurate Is AI Video Analytics? Test It, Don't Trust It September 13, 2026 12 min. read # How Accurate Is AI Video Analytics? Test It, Don't Trust It AI & Security AI Video AnalyticsComputer VisionPerimeter SecurityPhysical Safety ![An industrial yard fence line at dusk in light rain, with a pole-mounted security camera angled along the fence and wet gravel reflecting sodium light.](/_astro/how-accurate-is-ai-video-analytics-testing-standard-hero.Bied7gjH_Z20Q2z.webp) A video analytics datasheet quotes an accuracy figure in the high nineties, and the figure cannot be wrong, because it never says what it was measured on. The buyer finds out what it means on their own fence line, on a wet night, when the alert queue fills with shadows or stays silent through the one event that mattered. The question “how accurate is AI video analytics?” has a real answer for a given site, but the percentage on the datasheet is not it. ## The problem: how accurate is AI video analytics, measured against what? An accuracy claim is only meaningful if four things come with it: the scenario being detected, the scene it was tested in, the conditions — light, weather, distance, camera movement — and which of two possible errors the number describes. Marketing figures for this category routinely arrive with none of the four. That is not necessarily dishonest. It is simply unfalsifiable, and a buyer cannot plan an operation around a number they cannot check. The deeper problem is arithmetic. In a security or safety scene, genuine events are rare, so a system can score almost perfectly while being useless. Take one camera watching a yard fence for 30 days, which is 43,200 minutes, with two genuine intrusions in that month. Score each minute as right or wrong. System What it does Minute-by-minute “accuracy” A Never alarms. Misses both intrusions. 43,198 of 43,200 minutes correct: **99.995%** B Catches both intrusions, plus 90 false alarms over the month 43,110 of 43,200 minutes correct: **99.79%** The system that would have caught both break-ins scores lower than the one that watches nothing. That is not a trick of the example. Any measure dominated by the long stretches where nothing happens will reward a detector for staying quiet. System B has its own problem, and the same table exposes it. Of its 92 alarms, two were real, so an operator who responds to every alert is right about one time in 46. A buyer who asks only “how accurate is it?” learns nothing about either system’s failure. > A detector that watches nothing can score 99.995%. ## Why the usual approach falls short Buyers are not naive about this. They have learned to discount datasheets, and they reach for three substitutes, each of which helps less than it appears to. **The vendor demo.** It runs on footage the vendor chose, in conditions the vendor knows the system handles. It proves the detection exists. It says nothing about how often it fires on your site’s birds, headlights and rain. **The pilot on the best camera.** A trial on the newest, best-placed camera in daylight is designed to succeed, and it usually does. The camera at the far corner of the yard, the one pointed into low sun at 5 p.m., is where the alerts will actually go wrong, and it was never in the test. **The request for a better number.** Asking for “false positive rate” instead of “accuracy” is progress, but a rate without a denominator repeats the original problem. False alarms per what — per camera, per hour, per event, per thousand frames? Each gives a different number for the same system on the same night. What is missing in all three is an external reference point: a way of stating a test that does not depend on trusting whoever ran it. That reference point exists, and it has existed for some time. ## What good looks like: test video analytics the way the standards do Video analytics accuracy is not one number but two, each stated for a named scenario and scene. Recall is the share of genuine events that produce an alarm. Precision is the share of alarms that correspond to a genuine event. A figure that does not say which it measures, or under what conditions, is a claim rather than a measurement. ### Two errors, two costs, and a role that decides between them A missed event and a false alarm are different failures, and which one hurts more depends on what happens to the alert. The UK Home Office made this explicit in its i-LIDS programme. The [i-LIDS User Guide](https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/143875/ilids-user-guide.pdf), published by its Centre for Applied Science and Technology in February 2011, scores systems on a weighted combination of recall and precision. It evaluates them for one of two roles. In the operational alert role, a human controller must deal with each detection in real time. In the event recording role, detections only trigger recording for later review, so false alarms matter less. The guide sets a separate weighting for each role and each scenario, and for abandoned baggage the recording role’s weighting towards recall was a hundred times the live-alert role’s. That is the most useful idea a buyer can take from the programme. Before asking how good a system is, decide what a false alarm costs your operators and what a missed event costs your site. If alarms go to The failure that costs most What to weight in your test A person who must respond now False alarms, which spend attention and train people to ignore the queue Precision, stated as false alarms per camera per day Recording or later review Missed events, which leave nothing to review Recall, stated as genuine events caught out of genuine events staged ### What a government test library taught the category i-LIDS paired that metric with discipline about the test itself. The [i-LIDS brochure](https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/143870/ilids-brochure.pdf) describes six scenarios, including abandoned baggage, doorway surveillance, parked vehicles and sterile zone monitoring, built from footage it says represents real operating conditions. Several design choices are worth copying directly. 1. **The certifying footage was private.** Each scenario had a public training set and a public test set, and certification ran on a third, private evaluation set held by the Home Office. 2. **Every alarm event was defined in advance.** An abandoned object, for example, counted only once its owner had been gone from the area for a set time. 3. **Timing was part of the answer.** A system had ten seconds from the start of an event to raise an alarm; anything later counted as a false positive. 4. **Noise was catalogued, not ignored.** For sterile-zone footage along fences, the records logged potential causes of false alarms — foxes, bats, insects on the lens, shadows through the fence, a camera switching from colour to monochrome — along with time of day, rain, snow and fog. 5. **The pass mark was never published.** The guide states that the scores needed for a recommendation were not made public, so a manufacturer could not tune to a known threshold. The programme dates from over a decade ago, and its hardware and formats show it. The method has not aged. ### What IEC 62676-6:2026 adds On 18 February 2026, the IEC published [IEC 62676-6:2026](https://webstore.iec.ch/en/publication/59704), a 224-page international standard for performance testing and grading of real-time intelligent video content analysis in surveillance systems. Its published scope says it specifies functions, performance, test methods, performance evaluation and grading rules, and that it applies to live and forensic analysis. The scope organises testing into three dimensions: * **Core capability** — classifying objects and detecting object activity such as stopping, starting and direction of movement. * **Complex capability** — detecting scenarios built from that activity, with examples including loitering, perimeter intrusion detection, person down, tailgating and abandoned object detection. * **Degree of difficulty** — testing under defined levels of operating stress, with examples such as sterile or non-sterile scenes, indoor or outdoor, target obscuration, extreme weather and camera shake. It also names end users, installers, integrators and certification providers among the people it is meant to give measurement methods to. A note on what this article can and cannot say. The full standard is sold, and everything above comes from its publicly readable scope. Nothing here describes its clauses, test procedures or grade boundaries, and no claim is made about any product’s grade. What matters for a buyer is that the vocabulary now exists in a published international standard: scenario, capability, and degree of difficulty. A vendor who claims performance can be asked to state it in those terms. ### Five questions that turn a percentage into a test Put these in the RFP, and ask for written answers. 1. **Which scenario, and what counted as a detection?** Name the event, when it starts, and how long the system had to alarm. 2. **Recall and precision, as counts.** How many genuine events were staged or present, how many were caught, and how many alarms fired in total, per camera per day. 3. **Under which conditions?** Day and night, weather, distance to target, obscuration and camera movement — the degree of difficulty, stated rather than implied. 4. **Was the test footage separate from the training footage?** A system scored on the footage it learned from has been marked on its own homework. 5. **Which role was it tuned for?** A live-alert configuration and a record-for-review configuration are different settings, and one set of figures cannot describe both. The [video analytics RFP questions vendors cannot bluff](/nsights/video-analytics-rfp-vendor-questions/) cover the architecture side of the same procurement: streams, edge processing and data retention. ### Writing an acceptance test for your own site No external standard can say what a system will do on your cameras. It can only tell you how to find out. A site acceptance test borrows the structure above and applies it to your own footage. Choose the two or three scenarios that justify the purchase. Include the worst cameras, not just the best: the long view, the backlit gate, the dock that floods with headlights at shift change. Run across nights and weather, not a single afternoon. Define each alarm event and the response window before the test starts, and count true alarms, false alarms and misses per camera per day. Keep the labeled log afterward: it is also how you would test a [decision model that ranks perimeter alerts by context](/nsights/decision-models-alert-fatigue-perimeter-security/). Then set the bar by role. If a guard must respond to every perimeter alert, agree the number of false alarms per camera per night the team can genuinely work, because the [cost of loitering and dwell-time false alarms](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/) is paid in attention long before it shows up anywhere else. If alarms only flag footage for review, weight the misses. ## Where Nsightify fits Nsightify does not publish an accuracy percentage for any detection, and after the arithmetic above it would be an odd thing to offer: a figure measured on someone else’s scene tells you little about yours. What Nsightify does is run its detections on the IP and CCTV cameras a site already operates, with real-time alerts, which means the test described above can be run on the real estate rather than a demonstration rig. Some detections are ready to run from day one, such as intrusion into a restricted zone, loitering, PPE compliance and forklift–pedestrian proximity. Others, such as perimeter pacing, fence-climbing and abandoned object detection, are included and switched on during commissioning once tuned against the site’s own footage, because a busy yard and a quiet one produce different results. The limits apply to Nsightify as to any system. Detection depends on camera placement, sightlines, lighting and weather, and the only honest answer to “how accurate is it here?” comes from counting alarms, misses and false alarms on your own cameras, in your own conditions. ## Frequently asked questions ### How accurate is AI video analytics? It depends on the scenario, the scene, the lighting and weather, and the alert threshold, which is why a single percentage cannot answer the question. Accuracy is two measurements: recall, the share of genuine events that trigger an alarm, and precision, the share of alarms that are genuine. Ask for both, as counts, for a named scenario under stated conditions, and confirm them on your own cameras. ### What is a good false positive rate for video analytics? There is no universal figure. The acceptable rate depends on what happens to an alarm. If a person must respond to every alert in real time, false alarms consume attention and should weigh heavily. If alarms only mark footage for later review, missed events cost more. Set the target as false alarms per camera per day, based on how many alerts your operators can genuinely work. ### Is there a standard for testing video analytics? Yes. IEC 62676-6:2026, published on 18 February 2026, specifies test methods and grading rules for real-time intelligent video analysis in surveillance systems. Its published scope covers object classification and activity, complex scenarios such as loitering and perimeter intrusion, and testing under defined operating stress. Before it, the UK Home Office’s i-LIDS programme set a public precedent for scenario-based testing against defined alarm events. ### Why does my camera keep sending false alarms? Usually because the detection is working in conditions it was never tested under. The UK government’s i-LIDS test footage for fence-line scenes logged false-alarm causes such as foxes, bats, insects on the lens, shadows through the fence and cameras switching between colour and monochrome, alongside rain, snow and fog. A system tuned on a clear daytime scene meets all of them at night. Test and tune on your own cameras. * * * ## What to change in your evaluation this week Strike the word “accuracy” from the vendor questionnaire and replace it with the five questions above. Pick the three cameras on your estate most likely to embarrass a detector, and write down, before any trial starts, what counts as an alarm event and how many false alarms per camera per day your team can work. The same camera list does double duty as the [stream inventory IT needs before approving video analytics](/nsights/video-surveillance-network-requirements-it-approval/), so it is not wasted if the trial moves forward. If you want to run that test on your own cameras rather than read about it, [apply to the Nsightify pilot program](/contact/?interest=pilot). More on this from Nsightify: [AI video analytics on existing IP and CCTV cameras](/solutions/). [Previous](/nsights/bipa-video-analytics-biometric-identifier/) [BIPA and Video Analytics: Where Biometric Collection Starts](/nsights/bipa-video-analytics-biometric-identifier/) [Next](/nsights/california-emotion-ai-workplace-law-ab-1883/) [California Emotion AI Law: The Definition Is Wider Than the Ban](/nsights/california-emotion-ai-workplace-law-ab-1883/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## AI Video Analytics Insights for Safety & Security Teams URL: https://nsightify.com/nsights/ Practical perspectives and deep-dives on AI video analytics, PPE compliance and perimeter security, from the Nsightify team. # AI Video Analytics Insights for Safety & Security Teams Field notes from the team building Nsightify: how AI video analytics runs on the IP and CCTV cameras a site already has, what PPE and hazard-zone detection catches in practice, and what changing safety regulation means for the people who have to comply with it. AI & Security Physical Safety News [Home](/) Nsights [![Wide view of a logistics and warehouse facility at dawn with workers walking toward an illuminated loading bay](/_astro/logistics-facility-operations-dawn.7RzeaFh-_T4Jag.webp)](/nsights/nsightify-pilot-program-founding-partner/) News AI Video AnalyticsCompany News News June 22, 2026 ## [The Nsightify Pilot Program Is Open — Become a Founding Partner](/nsights/nsightify-pilot-program-founding-partner/) A limited pilot for a small group of operations and security teams — real-time intelligence on your existing cameras. Founding partner spots are limited. [![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_24cE1X.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/decision-models-alert-fatigue-perimeter-security/) AI & Security Perimeter Security By Ricardo Assaad October 7, 2026 ## [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [Read More](/nsights/decision-models-alert-fatigue-perimeter-security/) [![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_hNi2E.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/video-analytics-at-night/) AI & Security AI Video Analytics October 6, 2026 ## [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [Read More](/nsights/video-analytics-at-night/) [![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__21U0E5.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/pixels-per-foot-video-analytics/) AI & Security Camera Integration October 1, 2026 ## [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) [Read More](/nsights/pixels-per-foot-video-analytics/) [![A dome security camera mounted above an empty warehouse loading dock with the dock doors closed and a painted floor line.](/_astro/employee-monitoring-laws-by-state-cameras-hero.DTXNao5P_Z3RfdN.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/employee-monitoring-laws-by-state-cameras/) Physical Safety Regulatory Compliance September 28, 2026 ## [Do Employee Monitoring Laws by State Cover Cameras?](/nsights/employee-monitoring-laws-by-state-cameras/) [Read More](/nsights/employee-monitoring-laws-by-state-cameras/) [![A trailer drop lot at dusk, rows of plain parked trailers behind a chain-link fence, with a pole-mounted camera and yard lights and no one in view.](/_astro/cargo-theft-statistics-2026-hero.B8pxRmKG_2jO9Ms.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/cargo-theft-statistics-2026/) AI & Security Perimeter Security September 25, 2026 ## [Cargo Theft Statistics 2026 Show Fewer, Bigger Thefts](/nsights/cargo-theft-statistics-2026/) [Read More](/nsights/cargo-theft-statistics-2026/) [![A worker in a high-visibility vest and hard hat seen from a distance down a warehouse aisle, small in frame and too far away for equipment detail to be legible.](/_astro/ppe-detection-what-cameras-cannot-verify-hero.CzOxMOBq_Z1qfad9.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/ppe-detection-what-cameras-cannot-verify/) Physical Safety PPE Compliance September 22, 2026 ## [PPE Detection: What a Camera Verifies, and What It Cannot](/nsights/ppe-detection-what-cameras-cannot-verify/) [Read More](/nsights/ppe-detection-what-cameras-cannot-verify/) [![A ceiling-mounted security camera on a steel roof truss inside a distribution warehouse, with racking and concrete floor out of focus below in flat daylight.](/_astro/california-emotion-ai-workplace-law-ab-1883-hero.BiT2xo-X_ZfwYoC.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/california-emotion-ai-workplace-law-ab-1883/) Physical Safety Regulatory Compliance September 21, 2026 Updated October 1 ## [California Emotion AI Law: The Definition Is Wider Than the Ban](/nsights/california-emotion-ai-workplace-law-ab-1883/) [Read More](/nsights/california-emotion-ai-workplace-law-ab-1883/) [![An industrial yard fence line at dusk in light rain, with a pole-mounted security camera angled along the fence and wet gravel reflecting sodium light.](/_astro/how-accurate-is-ai-video-analytics-testing-standard-hero.Bied7gjH_96orp.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/how-accurate-is-ai-video-analytics-testing-standard/) AI & Security AI Video Analytics September 13, 2026 ## [How Accurate Is AI Video Analytics? Test It, Don't Trust It](/nsights/how-accurate-is-ai-video-analytics-testing-standard/) [Read More](/nsights/how-accurate-is-ai-video-analytics-testing-standard/) [![An empty industrial employee entrance corridor with a blank wall-mounted time clock beside a steel door and a small dome camera in the ceiling corner.](/_astro/bipa-video-analytics-biometric-identifier-hero.DQVqytK5_ZfvWXy.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/bipa-video-analytics-biometric-identifier/) Physical Safety Regulatory Compliance September 12, 2026 ## [BIPA and Video Analytics: Where Biometric Collection Starts](/nsights/bipa-video-analytics-biometric-identifier/) [Read More](/nsights/bipa-video-analytics-biometric-identifier/) [![A long internal corridor in an industrial facility, with a wall-mounted security camera in the foreground angled away down the corridor past two plain closed doors.](/_astro/california-workplace-surveillance-camera-placement-hero.DWkt5kaB_cGddy.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/california-workplace-surveillance-camera-placement/) Physical Safety Regulatory Compliance September 8, 2026 Updated October 1 ## [California Workplace Surveillance Laws: Where Cameras Cannot Go](/nsights/california-workplace-surveillance-camera-placement/) [Read More](/nsights/california-workplace-surveillance-camera-placement/) [![A network equipment room in an industrial facility, with a wall-mounted patch panel of dressed ethernet cabling and an annotated printed floor plan open on a table in the foreground.](/_astro/video-surveillance-network-requirements-it-approval-hero.DOujAaUC_ZuTC8E.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/video-surveillance-network-requirements-it-approval/) AI & Security Camera Integration September 7, 2026 ## [Video Surveillance Network Requirements: Getting IT to Yes](/nsights/video-surveillance-network-requirements-it-approval/) [Read More](/nsights/video-surveillance-network-requirements-it-approval/) [![An industrial commercial facility perimeter fence line and automated gate at dusk with mounted security cameras and crisp exterior lighting.](/_astro/loitering-vs-dwell-time-perimeter-false-alarms-hero.ChCVkXbe_228UxL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/) AI & Security Perimeter Security September 1, 2026 Updated October 8 ## [Loitering Detection False Alarms: Calibrating Dwell Time](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/) [Read More](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/) [![An automated industrial robotic cell in a modern manufacturing plant with high-visibility yellow perimeter floor markings and overhead surveillance cameras, under crisp natural facility lighting.](/_astro/machine-guarding-restricted-zone-optical-detection-hero.B9YzRqHS_Z2wNKNo.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/machine-guarding-restricted-zone-optical-detection/) Physical Safety OSHA August 31, 2026 ## [Machine Guarding Restricted Area: Where Optical Detection Fits](/nsights/machine-guarding-restricted-zone-optical-detection/) [Read More](/nsights/machine-guarding-restricted-zone-optical-detection/) [![An enterprise edge computing server appliance mounted in an organized IT equipment rack inside a commercial facility, with neatly routed network cables and soft background view of an operations floor.](/_astro/video-analytics-rfp-vendor-questions-hero.Ki1Pnk1f_Z1z6Bx1.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/video-analytics-rfp-vendor-questions/) AI & Security AI Video Analytics August 25, 2026 ## [Video Analytics RFP Questions: 9 Tests Vendors Cannot Bluff](/nsights/video-analytics-rfp-vendor-questions/) [Read More](/nsights/video-analytics-rfp-vendor-questions/) [![An empty shift supervisor's desk at the edge of a distribution center floor, with a two-way radio on its charger and a clipboard, in flat daylight.](/_astro/california-ai-workplace-law-sb-947-hero.DpbRGsM5_Z1HDo7b.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/california-ai-workplace-law-sb-947/) Physical Safety Regulatory Compliance August 17, 2026 Updated October 1 ## [California AI Workplace Law SB 947: A Human Must Corroborate](/nsights/california-ai-workplace-law-sb-947/) [Read More](/nsights/california-ai-workplace-law-sb-947/) [![An unbranded dome security camera in the foreground of a warehouse corridor, with yellow floor markings, stacked pallets and an open roller door at the far end.](/_astro/ndaa.BWV7quXu_Z2aP2SU.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/ndaa-compliant-cameras-fcc-import-ban/) AI & Security Regulatory Compliance August 12, 2026 ## [NDAA Compliant Cameras: You May Not Need to Replace Yours](/nsights/ndaa-compliant-cameras-fcc-import-ban/) [Read More](/nsights/ndaa-compliant-cameras-fcc-import-ban/) [![Empty loading dock aisle with a marked pedestrian walkway and an illuminated exit route sign overhead.](/_astro/osha-warehouse-emphasis-program-2026-hero.dIFz7tYk_Z1BhNQc.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/osha-warehouse-emphasis-program-2026/) Physical Safety OSHA August 5, 2026 ## [OSHA's Warehouse National Emphasis Program Now Runs to 2031](/nsights/osha-warehouse-emphasis-program-2026/) [Read More](/nsights/osha-warehouse-emphasis-program-2026/) [![Close-up of an existing IP security camera mounted on a concrete wall at an industrial facility](/_astro/existing-ip-camera-closeup.DYAOVL39_Z1OukAf.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/you-dont-need-new-cameras-ai-video-analytics/) AI & Security AI Video Analytics June 22, 2026 ## [You Don't Need New Cameras to Get AI Video Analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) [Read More](/nsights/you-dont-need-new-cameras-ai-video-analytics/) [![Worker wearing a hard hat and high-visibility vest walking through a manufacturing floor with an AI-detected hazard zone highlighted](/_astro/worker-ppe-hard-hat-manufacturing-floor.Da9WuQaA_F0g13.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/physical-safety-monitoring-real-time-ai/) Physical Safety AI Video Analytics June 20, 2026 ## [Real-Time Physical Safety Monitoring on Cameras You Own](/nsights/physical-safety-monitoring-real-time-ai/) [Read More](/nsights/physical-safety-monitoring-real-time-ai/) [![Security cameras mounted on an industrial perimeter fence overlooking a logistics yard at dusk, with AI data visualization overlay](/_astro/security-cameras-industrial-site-ai.D30p1adF_Z20kvUe.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ ](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) AI & Security AI Video Analytics June 19, 2026 ## [AI Video Analytics: Turn Existing Cameras Into Intelligence](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) [Read More](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) Show 10 more articles #nsights-posts\[data-collapsed\] .mxd-post.is-overflow { display: block } .nsights-showmore { display: none } Categories * AI & Security * Physical Safety * News Stay updated Practical perspectives on AI video analytics, physical safety, and perimeter security — from the team building Nsightify. Done! Thanks for subscribing. Something went wrong. Please try again. About Practical perspectives on AI video analytics, physical safety, and perimeter security — from the team building Nsightify. See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Loitering Detection False Alarms: Calibrating Dwell Time URL: https://nsightify.com/nsights/loitering-vs-dwell-time-perimeter-false-alarms/ Loitering detection false alarms start with one timer that treats every pause as a breach. How zones, object classes and schedules cut the noise. [Home](/) [Nsights](/nsights/) Loitering Detection False Alarms: Calibrating Dwell Time September 1, 2026 Updated October 8, 2026 12 min. read # Loitering Detection False Alarms: Calibrating Dwell Time AI & Security Perimeter SecurityAI Video AnalyticsCamera IntegrationComputer Vision ![An industrial commercial facility perimeter fence line and automated gate at dusk with mounted security cameras and crisp exterior lighting.](/_astro/loitering-vs-dwell-time-perimeter-false-alarms-hero.ChCVkXbe_Z2ogH12.webp) A delivery driver steps out of a cab at an entrance gate to check shipping paperwork on a clipboard. Forty seconds later, an automated alarm flashes on the monitoring console in the security operations center, flagging an unauthorized intruder lingering on the perimeter. By the fourth time this happens during a morning shift, the security officer clicks acknowledge without looking at the video feed. When an intruder spends two minutes testing fence fabric behind a row of trailers, that alert lands in the same noisy queue. Behind most loitering detection false alarms sits one setting: a static timer that treats every pause as a breach. ## Where loitering detection false alarms come from Loitering detection is a video analytics rule that raises an alert when a person or vehicle stays inside a drawn zone longer than a set time. Most systems implement it as one timer per zone: an installer draws a box around a gate, yard or fence line and picks a threshold, such as thirty seconds. Everything the rule knows is how long something has stayed. In a sterile test environment, that logic holds. In a live commercial facility, it collapses under normal daily operations. False Alarm Mechanics  /  Dwell Baselines Commercial perimeter dwell realities & noise Routine operations vs. hostile reconnaissance 4 Real-world dwell scenarios 01 Inbound Logistics False Loitering Alarm ### Delivery Driver Paperwork Dwell Pattern: Stationary 45s at Gate Driver checking manifests and gate pass on clipboard. Legacy 30s timer flags routine delivery as active intrusion. Nuisance alarm · Dispatches guards unnecessarily 02 Operational Staff False Loitering Alarm ### Guard Shift Changeover Dwell Pattern: Stationary 60s at Gatehouse Security officers debriefing during shift handover. Static box timer triggers repeatedly during daily turnover. Nuisance alarm · Trains staff to ignore alerts 03 Freight Transit False Loitering Alarm ### Staged Tractor-Trailer Dwell Pattern: Stationary 120s in Yard Truck waiting for spotter or dock door assignment. Pixel motion triggers continuous false loitering warnings. Nuisance alarm · Clutters SOC console queue 04 Critical Threat Missed in Noise ### Hostile Fence Inspection Dwell Pattern: Pacing / Dwell along Fence Intruder scoping fence fabric, locks, and camera angles while continuously moving. Lost in queue noise. Critical breach risk · Delayed or zero response When static timers treat every pause as a breach, real perimeter reconnaissance gets lost in console alarm noise. Illustrative scenarios · Example dwell times The intrusion alarm industry has already lived through this problem at scale. In its problem-oriented policing guide [False Burglar Alarms](https://portal.cops.usdoj.gov/resourcecenter/content.ashx/cops-p014-pub.pdf) (2nd edition, August 2011), the U.S. Department of Justice Office of Community Oriented Policing Services reports that between 94 and 98 percent of the alarm calls police respond to are false, higher in some jurisdictions. The guide gives no single measurement year; the examples beside the figure are U.S. city records from 1999 to 2004. That figure describes panel-based burglar alarms rather than camera analytics, and the mechanism is the one that matters here: a detector with no way to judge context produces alerts nobody can afford to trust. The industry’s answer at the panel layer was [ANSI/SIA CP-01](https://www.securityindustry.org/industry-standards/cp-01-2019/), the Control Panel Standard – Features for False Alarm Reduction, which recommends design features for control panels and the devices that arm and disarm them. It governs how a panel arms, disarms and reports. Nothing in it tells an outdoor camera what should count as suspicious in the first place. The cost lands on the person watching the console. Every nuisance alert spends a unit of attention, and attention is finite across a twelve-hour shift. Once the queue is mostly noise, acknowledging without watching becomes the rational response. Security teams call this [alert fatigue](/nsights/decision-models-alert-fatigue-perimeter-security/). Two fixes tempt from here, and both fail. Raising the timer to five minutes quiets the console and hands an intruder five minutes. Switching the loitering alert off quiets it completely. ## Three assumptions built into a single loitering timer One timer per zone quietly assumes three things: that every zone has the same normal dwell, that direction doesn’t matter, and that every object deserves the same rule. Each assumption fails somewhere different. ### 1\. The assumption of uniform dwell across diverse operational zones A single facility contains distinct operational environments with very different dwell baselines. At a typical distribution site, for example: * At an **access gate**, a truck waiting for clearance idles for a minute or more. * In a **staging yard**, a forklift driver setting down a pallet stays put for thirty seconds before reversing. * At an **external employee break area**, workers stand together for ten to fifteen minutes. * Along an **isolated rear fence line**, a person standing still for more than fifteen seconds has no operational reason to be there. When a security system applies one thirty-second timer across every outdoor camera, it raises a stream of nuisance events in working zones while failing to provide sensitive, early detection where exposure is highest. ### 2\. Missing directional and vector intelligence Legacy loitering analytics track bounding boxes without calculating directional intent. A person walking on a public sidewalk adjacent to a facility fence line enters the camera field of view and moves steadily at three miles per hour. Because the camera has a wide-angle lens covering a long perimeter span, the pedestrian remains inside the detection zone for thirty-five seconds simply by walking past. Spatial Vector Analysis  /  Trajectory Tracking Pedestrian transit velocity vs. hostile loitering vectors Linear momentum vs. spatial reversals Suppression without blind spots Benign Pedestrian Transit (Public Right-of-Way) Continuous linear velocity · Straight path along public sidewalk Suppressed (Normal) 01 Transit Sidewalk Entry Ingress Vector Pedestrian enters camera view on public sidewalk at steady forward walking pace. 02 Transit Linear Velocity Constant Momentum Maintains 3+ MPH vector parallel to perimeter fence with zero pauses or reversals. 03 Transit Perimeter Exit High Net Displacement Clears camera field smoothly within expected window; suppresses nuisance alert. VS HOSTILE RECONNAISSANCE Hostile Perimeter Loitering & Reconnaissance Directional reversals · Low net spatial displacement near barrier Instant Alert 01 Loitering Perimeter Approach Vector Divergence Turns off public sidewalk and heads directly toward fence line or access gate. 02 Loitering Stationary Pause Velocity Drop Halts motion within 5 feet of physical barrier exceeding calibrated dwell limit. 03 Loitering Directional Pacing Vector Reversals Walks back and forth repeatedly checking camera angles and physical locks. 04 Loitering Barrier Inspection Low Net Displacement Probes fence fabric and gate latches; raises an immediate alert. Vector analytics evaluates path linearity, velocity, and spatial reversals—suppressing sidewalk foot traffic while flagging fence probing. Illustrative example · Two paths through one zone Without spatial vector analysis, the camera cannot distinguish between a pedestrian walking in a straight line past the property and an individual pacing back and forth to inspect security gates. Both trigger the same loitering rule because both stayed in the frame for more than thirty seconds. ### 3\. Inability to classify object types and staging states Basic pixel-based motion detection has no idea what an object is. When a logistics contractor parks a flatbed trailer along a yard perimeter, the detector treats the trailer as a new object. As shadows shift or headlights sweep across the trailer body, the changing pixels register as activity, and the alerts repeat through the night. True perimeter visibility requires classifying people, passenger vehicles, and heavy commercial trucks independently, applying distinct temporal rules to each classification. > An alarm system that alerts on every delivery truck teaches your security team to ignore real perimeter breaches. ## Directional dwell: rules by zone, object class and time of day Directional dwell replaces the single timer with rules that ask more than one question. Instead of asking how long an object has been in a box, the system evaluates what the object is, where it entered, which direction it is traveling, and whether its dwell pattern matches legitimate operations. Concentric Defense  /  Perimeter Architecture Multi-layered perimeter zone architecture Concentric exposure zones Calibrated temporal thresholds Zone 03 Outer Transit Buffer ### Public Right-of-Way Buffer (Sidewalk & Roadway Approach) Vector Tracking → Monitored for directional approach vectors and pacing toward fence; suppresses parallel pedestrian and vehicle transit. System Action: Directional approach alerts · Suppresses parallel transit false alarms Zone 02 Sterile Security Boundary ### Exterior Fence Line Buffer (10-Foot Sterile Strip Along Fence) Tripwire & 15s Dwell → Monitored for boundary crossing, fence climbing, and low-velocity human dwell exceeding 15 seconds near physical barriers. System Action: High-priority alert · Sent to the on-site team at once Zone 01 Operational Facility Core ### Active Yard & Gate Operations (Loading Docks, Gates & Staging Yards) Multi-Threshold Dwell → Contextual multi-threshold dwell (up to 180s for commercial trucks), tightening to zero-tolerance intrusion rules after hours. System Action: Accommodates manifest review · Zero-tolerance intrusion after hours Concentric zoning applies tight 15-second thresholds to sterile fence lines while granting operational tolerances to active gates. Illustrative configuration · Example thresholds Directional dwell analytics evaluates a classified person or vehicle inside a defined perimeter zone by its trajectory, its travel speed and how long it stays. It separates authorized operational activity from stationary loitering or perimeter reconnaissance without relying on one timer for the whole site. Evaluation Dimension Static Bounding-Box Timers (Legacy CCTV) Directional Dwell Analytics (Modern Computer Vision) **Primary Trigger** Total elapsed seconds inside a 2D bounding box Trajectory, velocity, object class, and dwell duration **Sidewalk & Gate Transit** Triggers false alarms on long camera spans Suppressed via linear vector continuity and transit velocity **Object Classification** Generic motion blob / unclassified bounding box Independent models for humans, light vehicles, and heavy trucks **Operational Staging** Floods console on parked trailers and break areas Handled via contextual time-of-day masks and multi-tier thresholds **Pre-Incident Detection** Blind to pacing or slow perimeter inspection Flags low net displacement and repeated perimeter passes The numbers in the four steps below are example starting points, not standards. Every site tunes its own against what its cameras actually see. ### Step 1: Establishing concentric perimeter buffer zones Concentric zoning is the structural fix. Instead of one detection box per camera, divide the site by exposure: the further a zone sits from legitimate daily operations, the tighter its dwell threshold. 1. **Outer Transit Zone (Public Right-of-Way):** Covers sidewalks and public access lanes. Configured with directional tripwires that track approach vectors toward the facility but ignore parallel foot and vehicular traffic. 2. **Perimeter Sterile Buffer (Fence Line):** A ten-foot interior buffer along fencing and property boundaries where no legitimate operational staging occurs. Configured with a tight fifteen-second dwell threshold and instantaneous crossing tripwires. 3. **Operational Core (Gates, Docks, and Staging Yards):** Configured with wider dwell tolerances during business hours, up to three minutes for a truck at the gate, automatically tightening to zero-tolerance intrusion detection after hours. ### Step 2: Calibrating multi-threshold timing by object class Different physical objects carry different operational baselines. Calibrated video analytics applies separate temporal rules based on neural network object classification: Object Classification  /  Temporal Logic Object-specific dwell matrix & classification rules Human vs. vehicle temporal rules Contextual false alarm suppression 01 Pedestrian Access 45s Threshold Object Class Classified Human Target Zone Inbound Gate Zone Alert Trigger Threshold: 45 Seconds Accommodates brief driver dismount, badge scan, or gatehouse intercom check. Extended loitering beyond 45s flags guard attention. Advisory notification · Gatehouse log 02 Logistics Transit 180s Threshold Object Class Classified Vehicle Target Zone Inbound Gate Zone Alert Trigger Threshold: 180 Seconds (3 Mins) Accommodates shipping manifest review, BoL paperwork inspection, and spotter coordination without nuisance alarms. Operational tolerance · Suppresses false alarms 03 Sterile Perimeter 15s Threshold Object Class Classified Human Target Zone Fence Line Buffer Alert Trigger Threshold: 15 Seconds Zero legitimate operational presence along exterior fencing. Stationary presence exceeding 15s triggers immediate priority escalation. Priority alert · Sent to the on-site team at once Neural network classification applies independent timers—granting 3 minutes to trucks while alerting on 15 seconds at the fence. Illustrative configuration · Example thresholds If a delivery truck stops at an inbound security gate, the system applies a three-minute vehicle threshold, allowing the driver to scan credentials and speak with the gate attendant without triggering an alarm. If a person exits that vehicle and walks toward an unmonitored equipment yard, a separate forty-five-second pedestrian rule activates immediately. ### Step 3: Detecting perimeter pacing and reconnaissance The behavior worth catching early doesn’t have to involve standing still. The Department of Homeland Security’s [Nationwide SAR Initiative indicators](https://www.dhs.gov/sites/default/files/2024-04/24_0424_NSI_SAR-Indicators-Behaviors-Tools-Analysts-Investigators%20-%202024_0.pdf) list testing or probing of security, and unusual or prolonged interest in a facility, among the behaviors analysts should look into. They also treat neither as criminal on its own, which is the same distinction a loitering rule has to draw. On a perimeter camera, the version a detection can look for is movement: walking a stretch of fence and back, stopping at a gate, returning to the same section. A single timer misses it, because the person never stays in one place long enough to trip it. Directional computer vision tracks the cumulative path within a zone. Repeated direction reversals, low net displacement over sixty seconds, or lingering within five feet of a barrier mark the pattern as perimeter pacing and reconnaissance. ### Step 4: Automating time-of-day sensitivity schedules Perimeter activity that is completely normal at 2:00 PM is worth an alert at 2:00 AM. During daytime operating shifts, yard cameras run wide dwell thresholds (two to three minutes) to accommodate material handlers, shift changes, and contractor staging. At the end of the operating shift, the rules switch to an after-hours schedule: * Dwell thresholds along all exterior zones compress to ten seconds. * Directional tripwires along gates switch to instant intrusion alerting. * Any classified person inside an equipment yard alerts the on-site team immediately. ## Running loitering and dwell rules on existing perimeter cameras Nsightify connects to the RTSP-compatible IP cameras already on the fence line, so nothing gets replaced. It deploys as Nsightify Cloud, or as On-Premises (Zero Trust) on your own servers, where no footage leaves your network. These detections are included in [Nsightify Perimeter Security](/solutions/perimeter-security/) and ready to run: * **Loitering:** someone stays in a monitored area past the dwell time you set. * **Restricted-area dwell:** the stricter rule for space where nobody should stand still. * **Directional tripwire crossing:** a line with a direction, so someone coming in raises an alert and everyone going out doesn’t. * **Stopped vehicle:** a vehicle stationary at a gate approach or in a fire lane longer than you allow. * **After-hours zone entry:** anybody in a scheduled zone, such as the yard at 2:00 AM. * **Camera tampering:** a lens blocked, spray-painted, defocused, or repositioned. Perimeter pacing and reconnaissance, wrong-way movement, and fence-climbing are included too, and switched on at commissioning once tuned against your own footage. Any detection can be put on a schedule: restrict it to a shift, a night window, or a weekend, and choose whether evidence is still captured outside it. Optical analytics has physical limits: [video analytics at night needs its own test](/nsights/video-analytics-at-night/), and [pixels per foot on the target](/nsights/pixels-per-foot-video-analytics/) decides how far along a fence a rule can work. Fog, heavy snow, and dirty domes degrade tracking. Nsightify sharpens the response of the people guarding a site. It is not a physical barrier. ## Frequently asked questions ### What is the technical difference between loitering and dwelling in video analytics? Dwelling measures any duration of classified human or vehicle presence inside a designated virtual zone regardless of intent, such as a truck waiting at an infeed gate. Loitering represents unauthorized or anomalous stationary presence that exceeds a calibrated operational threshold or occurs in restricted zones during non-operational hours without legitimate transit activity. ### Why do simple 30-second camera loitering timers generate so many false alarms? Static timers trigger on total elapsed time inside a bounding box without evaluating direction, velocity, or operational context. Delivery drivers checking manifests, employees in designated outdoor break areas, and contractors staging equipment all stay in place longer than 30 seconds, so each of them trips the timer. ### How does directional dwell filtering reduce false alarm rates along fence lines? Directional dwell tracks the entry vector and vector continuity of people and vehicles. Pedestrians walking parallel to a fence line along an exterior public sidewalk maintain forward momentum and clear the field of view, while individuals pacing back and forth, changing direction repeatedly, or lingering near structural weak points trigger immediate investigation alerts. ### Does deploying calibrated loitering analytics require installing new IP cameras? No. Video analytics platforms read the stream an existing IP camera already produces, usually over RTSP, and older analog CCTV cameras can be reached through the recorder or an encoder that serves the same kind of stream. Directional tripwires, polygon masks, and multi-threshold dwell rules are configured in software, without replacing camera hardware or running new cabling. ### How do you evaluate false positives in loitering and intrusion detection? Count them against a baseline from your own site. Log every loitering and intrusion alert for two to four weeks, and have a reviewer mark each one as a real event or a nuisance, with its cause: a delivery check-in, sidewalk traffic, a parked trailer. Report nuisance alerts per camera per shift, then change one threshold or zone at a time and measure again over the same length of time. * * * ## Start with your five noisiest perimeter cameras Cutting loitering false alarms doesn’t require new cameras or more people watching them. Start with the five exterior cameras that alert most, typically the main logistics gate, a contractor entrance, and the primary yard fence lines. Take three calibration steps on each camera: 1. **Find your top three nuisance triggers:** Review the last thirty days of alarm logs. Identify whether delivery check-ins, shift changes, or sidewalk foot traffic caused most of the nuisance alerts. 2. **Separate pedestrian and vehicle thresholds:** Replace one timer per zone with thresholds by object class, for example forty-five seconds for people and three minutes for trucks at the gate. 3. **Draw the zone along your property line:** Mask public sidewalks and roadway lanes out of the loitering zone, so the rule watches only your side of the fence. Masking narrows what the rule analyzes. The camera still records the sidewalk, and the recorder still keeps that footage. Then count again over the same thirty days, the way [video analytics accuracy should be tested](/nsights/how-accurate-is-ai-video-analytics-testing-standard/): on your own footage, against the baseline you just made. When the queue holds fewer, better alerts, the officer who used to click acknowledge starts watching again. To watch dwell rules drawn on your own gate and fence views, [schedule a perimeter security demo with Nsightify](/contact/?interest=perimeter-security). More on this from Nsightify: [AI video analytics on existing IP and CCTV cameras](/solutions/). [Previous](/nsights/machine-guarding-restricted-zone-optical-detection/) [Machine Guarding Restricted Area: Where Optical Detection Fits](/nsights/machine-guarding-restricted-zone-optical-detection/) [Next](/nsights/video-surveillance-network-requirements-it-approval/) [Video Surveillance Network Requirements: Getting IT to Yes](/nsights/video-surveillance-network-requirements-it-approval/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Machine Guarding Restricted Area: Where Optical Detection Fits URL: https://nsightify.com/nsights/machine-guarding-restricted-zone-optical-detection/ OSHA 1910.212 requires physical machine guards. Here is how optical zone detection on existing cameras alerts on perimeter creep before an incident occurs. [Home](/) [Nsights](/nsights/) Machine Guarding Restricted Area: Where Optical Detection Fits August 31, 2026 11 min. read # Machine Guarding Restricted Area: Where Optical Detection Fits Physical Safety OSHAWarehouse OperationsRegulatory Compliance ![An automated industrial robotic cell in a modern manufacturing plant with high-visibility yellow perimeter floor markings and overhead surveillance cameras, under crisp natural facility lighting.](/_astro/machine-guarding-restricted-zone-optical-detection-hero.B9YzRqHS_Z1XmJXw.webp) A maintenance technician steps over a painted yellow boundary line to clear a minor cardboard jam on an active packaging line. If the interlock is bypassed or the safety light curtain was never reached, the facility’s only record of the near-miss is the one someone chooses to write down. The gap between mechanical guarding at the point of operation and supervisory visibility across the wider work cell is where severe industrial injuries happen. Physical machine guards are engineered to stop mechanical motion when a barrier opens or a light curtain breaks. What they cannot do is tell an operations manager that workers routinely step inside the hazard perimeter during active cycles, that an interlock latch was defeated with a zip tie, or that a forklift staged material within the swing radius of a robotic arm. Understanding where optical detection fits alongside federal safety mandates transforms machine safety from a reactive compliance checklist into continuous operational oversight. ## The problem with relying solely on physical machine guards Federal regulations governing machinery hazards are among the most strictly enforced standards in general industry. Under the Occupational Safety and Health Administration standard [29 CFR 1910.212(a)(1)](https://www.osha.gov/laws-regs/regulations/standardnumber/1910/1910.212), employers must provide one or more methods of machine guarding to protect operators and other employees from hazards created by point of operation, ingoing nip points, rotating parts, flying chips, and sparks. Despite decades of regulatory enforcement, machine guarding failures remain a persistent operational crisis. According to published enforcement data from OSHA, standard 1910.212 consistently ranks among the top ten most frequently cited workplace safety violations nationwide, generating thousands of citations and millions of dollars in penalties across manufacturing and warehousing sectors every year. OSHA 1910.212  /  Facility Exposure Map Industrial machine hazard exposure & blind spots Point of operation vs. perimeter buffer 3 Distinct hazard envelopes 01 Active Safeguarding 1910.212 Compliant ### Point of Operation Safeguard Level: Physical Guards & Interlocks Covered by OSHA 1910.212 barrier guards and hardwired E-Stop circuits. Halts motion when physical contact occurs. Binary trip only · Zero early behavioral warning 02 Critical Operational Gap Operational Blind Spot ### Perimeter Hazard Zone Safeguard Level: Unmonitored Machine Envelope Unmonitored until a physical breach occurs. Workers routinely step inside perimeter lines or defeat interlocks during active cycles. Near-misses unrecorded · Primary amputation risk 03 Dynamic Traffic Risk High-Traffic Transit ### Material Staging Buffer Safeguard Level: Forklift & Pedestrian Cross-Traffic High-traffic loading aisles and pallet staging lanes. Shared space creates recurring blind-corner and crush hazards. Unstructured traffic flow · High collision frequency Physical guards protect the point of operation; perimeter creep and transit congestion remain unmonitored blind spots. Source: OSHA 29 CFR 1910.212 · Machine Guarding Risk Analysis The human toll behind those citations is specific. OSHA’s [Machine Guarding eTool](https://www.osha.gov/etools/machine-guarding) states that workers who operate and maintain machinery suffer approximately 18,000 amputations, lacerations, crushing injuries, and abrasions, plus over 800 deaths, per year. The underlying problem is not that physical guards fail to work when installed properly. The problem is that physical guarding operates under a binary mechanical assumption: either the guard is intact and closed, or the emergency circuit trips and the machine halts. Physical guards provide zero telemetry about the human behaviors, staging bottlenecks, and operational shortcuts that precede a catastrophic breach. ## Why the usual approach falls short Most industrial plants manage machinery safety through three traditional mechanisms: fixed barrier guards, presence-sensing safety devices (such as optical light curtains and pressure mats), and periodic supervisory walkthroughs. While each serves a distinct engineering function, relying on them as the sole safety oversight mechanism creates major operational vulnerabilities. ### 1\. The interlock defeat paradox When an emergency stop or safety light curtain trips, the machine halts abruptly. In high-speed automated packaging, stamping, or palletizing cells, an unexpected line stoppage causes product jams, requires manual indexing, and creates production downtime. When operators face aggressive shift quotas, the friction of repeated line stops creates an incentive to defeat safety devices. Across fabrication plants and distribution hubs, workers tape over interlock switches, misalign optical sensors, or step around light curtains using structural framing to clear jams without cutting machine power. A physical guard provides no operational alert when it is being circumvented; it simply sits inactive while production continues. ### 2\. The limitation of point-of-operation presence sensing Under consensus standards such as [ANSI B11.19-2019](https://www.b11standards.org/storeitems/ansi-b1119) (_Performance Requirements for Risk Reduction Measures: Safeguarding and other Means of Reducing Risk_), presence-sensing devices are engineered around strict stopping-time formulas. The distance between a light curtain and the hazard must be calculated based on the maximum approach speed of a human hand or body relative to the mechanical stopping time of the machine. While ANSI-compliant light curtains protect the exact point of operation, they monitor only a narrow optical plane. They do not monitor the broader restricted envelope surrounding the machinery. If an unauthorized worker enters the perimeter staging zone behind an operator, or if a forklift mast intrudes into the overhead travel zone of an automated gantry, the light curtain remains completely unaware until an impact occurs. ### 3\. Periodic walkthroughs capture static compliance, not dynamic risk Safety audits and supervisor walkthroughs inspect machines in a static state. An auditor checks whether guards are bolted into place, whether emergency stop buttons are functional, and whether warning placards are visible. What a walkthrough cannot capture is shift-dependent behavioral drift. Walkthroughs miss the third-shift habit of propping open a maintenance gate to speed up changeovers, the temporary placement of scrap bins inside a robot’s operational radius, or the routine entry of material handlers into designated machinery clearance paths. Because these behaviors occur intermittently, they remain invisible to management until an OSHA inspection or an injury investigation takes place. ## What good looks like: Layered machine zone safety Defensible industrial safety requires separating physical machine safeguarding from operational zone monitoring. The two systems do not compete; they operate at different layers of the safety architecture. Safety Architecture  /  Layered Defense Model Three-layer machine safety architecture Concentric hazard envelopes Physical barriers + Optical AI Layer 03 Outer Transit Buffer ### Floor Buffer Zone (Pedestrian & Vehicle Transit) Traffic Separation → Monitored by overhead cameras for forklift–pedestrian traffic separation and clear egress paths System Action: Non-disruptive routing guidance · Staging violation alerts Layer 02 Predictive Observation Envelope ### Restricted Perimeter (Machine Envelope Buffer (3–6 ft Radius)) Optical AI Monitoring → Monitored by Optical AI for unauthorized dwell, boundary crossing, and bypassed interlocks System Action: Real-time supervisor alerts · Empirical near-miss audit log · Zero line stops Layer 01 Physical Mechanical Core ### Point of Operation (Physical Hazard Point (Nip Points, Dies & Robots)) OSHA 1910.212 & ANSI B11.19 → Protected by OSHA 1910.212 physical barriers, interlocked access gates, and certified ANSI E-Stops System Action: Hardwired circuit break · Halts machine motion immediately upon physical breach Optical detection operates at Layers 2 & 3 as an early observation layer without altering certified Layer 1 machine control circuits. Source: ANSI B11.19-2019 · Concentric Machinery Safeguarding Model Optical machine zone detection is the automated use of computer vision algorithms on existing overhead cameras to observe, classify, and timestamp unauthorized human or vehicle entry into designated virtual hazard perimeters around industrial equipment without altering machine electrical controls. By establishing virtual perimeter boundaries around automated equipment, facilities create an early-warning observation layer that catches safety encroachments before a worker reaches the mechanical hazard. Evaluation Dimension Active Safety Interlocks (ANSI B11.19 / OSHA) Optical Zone Monitoring (Computer Vision) **Primary Purpose** Physical prevention of contact at point of operation Operational visibility, near-miss logging, and early warning **System Action** Hardware-level electrical circuit break / Emergency stop Real-time supervisory alert (SMS, dashboard, webhook) **Hardware Boundary** Hardwired into machine safety PLC / control relay Connects to existing overhead IP/CCTV camera RTSP feeds **Reaction Horizon** Milliseconds (mechanical stopping distance) Real-time observation of approach and dwell behaviors **Audit Capability** Binary trip counter (often unrecorded) Structured metadata, timestamps, duration, and visual review **Operational Impact** Halts production immediately upon breach Delivers early alerts without triggering line stoppages ### Step 1: Mapping the three concentric safety zones A reliable machine vision safety architecture defines three distinct geometric zones around every automated or high-consequence asset: 1. **Zone 1: The Point-of-Operation Hazard Core.** The physical envelope containing nip points, cutting heads, stamping dies, or robotic arms. This zone is strictly protected by OSHA-compliant physical enclosures, interlocked gates, and ANSI B11.19 light curtains. 2. **Zone 2: The Virtual Restricted Perimeter.** A virtual polygon drawn around the machine footprint, extending three to six feet beyond the physical guard perimeter. This zone marks the boundary where only authorized operators should enter during active machine cycles. 3. **Zone 3: The Material Handling and Transit Buffer.** The wider staging lane where forklifts, tuggers, and material handlers stage raw goods or remove finished pallets. ### Step 2: Configuring contextual dwell and direction logic Simple motion detection triggers an alert whenever any pixel changes in the frame, flooding supervisors with false alarms from vibrating machinery, flickering indicator lights, or passing shadows. Defensible computer vision uses object classification and spatial tracking to evaluate human and vehicle behavior contextually. Spatial Tracking  /  Detection Logic Virtual polygon configuration & dwell logic Spatial tracking + dwell classification Zero false alarm architecture Physical-to-Virtual Ingress Sequence 01 Transit Zone Conveyor Infeed Material Staging Lane Unrestricted material buffer where forklifts & tuggers drop pallets. 02 Active AI Zone Virtual Polygon (Zone 2) Predictive AI Perimeter Dynamic 3–6 ft envelope monitored by computer vision spatial tracking. 03 Hazard Core Balers / Presses (Zone 1) Mechanical Hazard Point Physical machine point of operation protected by hardwired interlocks. Zone 2 Spatial Logic Evaluation Authorized Operator Normal Operation Trigger Condition: Dwell > 10s · Assigned Workstation Position Operator actively loading infeed conveyor, clearing staging area, or conducting scheduled changeover. System verifies continuous permitted presence without false alarm dispatch. System Action: Zero alarm · Structured telemetry logged to shift summary Unauthorized Transit Instant Alert Trigger Condition: Cross-Line Encroachment · Transient Shortcut Path Pedestrian cutting through machine buffer zone to bypass main aisle or reaching past yellow perimeter. Directional tripwire detects encroachment immediately upon entry. System Action: Instant SMS/Webhook to floor supervisor · Visual strobe alert Spatial AI evaluates dwell duration, direction, and posture—distinguishing normal operations from dangerous shortcuts in real time. Nsightify Computer Vision Pipeline · Spatial Dwell & Tripwire Engine In an effective configuration: * **Instantaneous Crossing Alerts:** Any unauthorized pedestrian crossing into Zone 2 while the equipment is operating triggers a high-priority alert. * **Directional Tripwires:** Alerting distinguishes between an operator stepping away from a workstation and a passerby cutting across a restricted maintenance lane. * **Dwell Thresholding:** Brief staging of materials at an infeed conveyor does not trigger an alarm, while extended human presence in a blind spot behind a baler generates immediate notification. ### Step 3: Connecting alerts to operational coaching, not line stops Optical zone monitoring does not wire into the machine’s electrical emergency stop circuit. This separation is deliberate. If camera software triggered an emergency stop every time an employee walked near a yellow line, the resulting production disruptions would lead operators to despise the system and find ways to obscure the cameras. Instead, optical detection dispatches real-time alerts directly to floor supervisors, area leads, or automated visual warning beacons (such as an amber perimeter strobe). A supervisor receives a notification with the exact zone, timestamp, and visual snippet, allowing them to intervene immediately, correct the unsafe behavior, and determine why the worker stepped inside the restricted envelope. ### Step 4: Building a defensible near-miss audit record In conventional safety programs, near-miss reporting relies on voluntary self-reporting. If a worker reaches into a running packaging machine to pull a loose strap and pulls their hand back safely, no report is filed. Management remains unaware that the hazard exists until an amputation occurs. Continuous optical monitoring logs every virtual boundary encroachment as structured metadata. Plant managers can review monthly trend data showing which machines experience the highest frequency of perimeter encroachments, which shifts account for the most boundary breaches, and whether specific equipment designs encourage shortcut behaviors. This empirical data is essential when reviewing compliance under [OSHA’s warehouse national emphasis program](/nsights/osha-warehouse-emphasis-program-2026/), giving EHS leaders verified documentation of proactive hazard identification and corrective coaching. ## Where Nsightify fits Nsightify provides an AI video analytics platform that adds operational visibility to the standard IP and CCTV cameras industrial facilities already operate. Our software connects directly to existing camera infrastructure over RTSP and ONVIF streams, requiring zero camera replacements, proprietary edge sensors, or hardware retrofits to machinery. Within our **Physical Safety** solution, Nsightify delivers allowlisted, ready-to-run detections purpose-built for industrial equipment environments: * **Entry into a restricted machine area:** Detects and timestamps unauthorized human presence inside defined virtual machine envelopes. * **Restricted-area dwell:** Monitors unauthorized lingering in blind spots, maintenance corridors, and hazardous equipment footprints. * **Unauthorized vehicle in a pedestrian zone:** Identifies forklift or tugger encroachment into machine operator walkways. * **Person down:** Immediately alerts on slips, falls, or sudden worker incapacitation near operating machinery. * **PPE compliance:** Verifies high-visibility vests, hard hats, and required protective equipment before personnel approach hazardous work cells. System Architecture  /  Zero-Trust Boundary Nsightify deployment boundary & isolation Zero changes to machine PLCs Passive RTSP stream ingestion 01 Source Layer Existing Fleet ### Existing IP Cameras Transport & Data: RTSP / ONVIF Profile S & T Connects to standard CCTV and IP cameras already mounted on facility columns and overhead trusses. 02 Compute Layer On-Premises AI ### Nsightify Edge Inference Transport & Data: Local Deep Learning Engine Processes video in volatile memory on site. Zero raw video streams leave the local network boundary. 03 Action Layer EHS Telemetry ### Supervisor Alerts & Audit Transport & Data: JSON Metadata & Webhooks Dispatches real-time SMS alerts, activates floor strobe beacons, and logs timestamped near-miss records. Non-Intrusive Safety Isolation Guarantee Zero PLC Impact Zero modifications to machine PLCs, electrical panels, control wiring, or certified physical E-Stops. Nsightify operates purely out-of-band as an operational intelligence layer over existing camera streams. It maintains total physical and electrical air-gap separation from safety-rated machinery control systems. Video inference runs on-premises over standard RTSP feeds without physical or electrical connection to machine controls. Source: Nsightify Physical Safety Architecture · Edge Deployment Model The operational boundaries must be stated with complete engineering honesty. Nsightify is a supervisory monitoring and alerting layer. It does not act as an emergency stop device (E-Stop), does not wire into machine safety relays, and does not replace certified OSHA 1910.212 physical barriers or ANSI B11.19 presence-sensing devices. Optical computer vision depends on camera sightlines, clean lenses, and adequate facility illumination. Blind spots behind heavy steel machine frames are a physical reality. That is precisely why Nsightify serves as an early-warning observation layer that strengthens [real-time physical safety monitoring on existing cameras](/nsights/physical-safety-monitoring-real-time-ai/) rather than attempting to substitute for mechanical safeguarding. Organizations can deploy Nsightify via Nsightify Cloud or as a Zero Trust on-premises appliance for air-gapped manufacturing environments where proprietary production layouts and video feeds must never leave the facility network. Understanding how [you don’t need new cameras to get AI video analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) allows plants to activate machine zone monitoring within days of commissioning. ## Questions plant managers and safety leaders are asking ### Does optical camera detection satisfy OSHA 1910.212 machine guarding requirements? No. OSHA 1910.212 and ANSI B11.19 require physical barriers or certified presence-sensing safety devices wired directly into machine control circuits. Optical camera detection serves as an operational monitoring layer that observes perimeter encroachment and logs near-misses, but it does not replace certified emergency stop mechanisms. ### What is the difference between a safety light curtain and optical zone monitoring? A safety light curtain is a hardware-interlocked safety device that immediately cuts power to a machine when an optical beam is broken at the point of operation. Optical zone monitoring uses standard overhead CCTV cameras and computer vision to detect, timestamp, and alert supervisors when personnel enter a wider restricted perimeter, without interrupting machine power. ### Can video analytics detect when a worker bypasses a physical machine guard? Yes. By establishing virtual restricted polygons around machinery envelopes, video analytics detects human presence inside hazardous zones regardless of whether physical interlocks were defeated, bypassed, or left open during operation, generating an immediate supervisory alert and an audit record. ### Does adding restricted machine zone monitoring require installing new cameras? No. Standard IP and CCTV cameras already mounted overhead or along facility structural columns provide sufficient resolution and field of view for computer vision algorithms to track virtual boundary crossings over standard RTSP streams. * * * ## What to audit around your equipment this week Improving machine safety does not require waiting for the next capital expenditure cycle or replacing operational machinery. Start by auditing the five highest-consequence automated machines on your floor—such as hydraulic balers, robotic palletizers, stretch wrappers, or metal stamping presses. Check three operational realities on each machine: 1. **Inspect physical interlocks for bypass evidence:** Look for magnetic overrides, taped sensors, or loose barrier latches. 2. **Observe actual operator clearance paths:** Identify where workers step during routine clearing of material jams and whether painted floor lines match daily workflows. 3. **Map camera coverage:** Verify whether existing overhead security cameras have an unobstructed line of sight across the machine’s restricted perimeter. By establishing virtual restricted zones on existing cameras, you create an objective, continuous record of perimeter encroachments that allows supervisors to address unsafe habits before they result in a recordable injury. If you are ready to add continuous zone monitoring to your plant floor, [talk to our team about Physical Safety monitoring](/contact/?interest=physical-safety). **Keep reading:** * [Real-time physical safety monitoring on cameras you own](/nsights/physical-safety-monitoring-real-time-ai/) * [OSHA’s warehouse national emphasis program now runs to 2031](/nsights/osha-warehouse-emphasis-program-2026/) * [You don’t need new cameras to get AI video analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) * [PPE detection and hazard-zone monitoring on existing cameras](/solutions/physical-safety/) More on this from Nsightify: [PPE detection and hazard-zone monitoring](/solutions/physical-safety/). [Previous](/nsights/video-analytics-rfp-vendor-questions/) [Video Analytics RFP Questions: 9 Tests Vendors Cannot Bluff](/nsights/video-analytics-rfp-vendor-questions/) [Next](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/) [Loitering Detection False Alarms: Calibrating Dwell Time](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## NDAA Compliant Cameras: You May Not Need to Replace Yours URL: https://nsightify.com/nsights/ndaa-compliant-cameras-fcc-import-ban/ The FCC's July 2026 rule bans imports of covered cameras, not their use. How to tell whether NDAA compliant cameras are actually required of you. [Home](/) [Nsights](/nsights/) NDAA Compliant Cameras: You May Not Need to Replace Yours August 12, 2026 14 min. read # NDAA Compliant Cameras: You May Not Need to Replace Yours AI & Security Regulatory ComplianceCamera IntegrationAI Video Analytics ![An unbranded dome security camera in the foreground of a warehouse corridor, with yellow floor markings, stacked pallets and an open roller door at the far end.](/_astro/ndaa.BWV7quXu_Zh2awz.webp) The FCC closed its legacy import loophole on 16 July 2026, and within a week the phrase NDAA compliant cameras was back at the top of every integrator’s quote. Nothing in that rule switches off an installed camera. Before you sign a replacement order, work out whether the obligation reaches you at all, and whether the project it is holding up actually needs to wait. ## What changed on 16 July 2026 The FCC’s Public Safety and Homeland Security Bureau and its Office of Engineering and Technology released [Public Notice DA 26-635](https://docs.fcc.gov/public/attachments/DA-26-635A1.pdf) on 26 June 2026. It was [published in the Federal Register](https://www.govinfo.gov/content/pkg/FR-2026-07-06/html/2026-13518.htm) on 6 July and took effect ten days later. What it does is narrow. Equipment added to the FCC’s Covered List in 2024 or earlier, and authorised before the November 2022 cut-off, can no longer be imported into the United States or marketed here. Until this notice, a model authorised before that cut-off could keep arriving indefinitely, which is why a decade-old design was still turning up on quotes in 2026. That gap is now closed. What it does not do is equally specific. The Commission revoked no equipment authorizations. It limited them. The machinery for that came from an order adopted on 29 October 2025 and codified at 47 CFR 2.939(e), which lets the Bureau stop future importation and marketing of covered equipment while the underlying authorization stands. The procedure requires a public notice and at least thirty days of comment before anything happens. The Federal Register notice then states the consequence directly: equipment you already possess, legally purchased and still holding an authorization, may continue in use. One boundary in the notice is easy to miss. It reaches equipment added to the Covered List in 2024 or earlier, and the list has kept growing since: foreign-produced uncrewed aircraft systems and the communications and video surveillance equipment named in section 1709 of the fiscal 2025 National Defense Authorization Act were added on 22 December 2025, routers on 23 March 2026, and power inverters and advanced robotic devices on 28 July 2026. None of those are covered by the July prohibition. The Bureau opened a separate proceeding for them the following day, in [Public Notice DA 26-742](https://docs.fcc.gov/public/attachments/DA-26-742A1.pdf) on 17 July 2026, proposing the same treatment and saying the same thing about use: importation and marketing would be prohibited, and equipment already in users’ hands would remain authorised. That one is a proposal out for comment, not a rule. _Current as of 12 August 2026. This is an open docket with follow-on proceedings; check the FCC’s Covered List page before relying on any date here._ ## What “NDAA compliant cameras” actually means NDAA compliant cameras are cameras containing no equipment produced by the manufacturers named in Section 889 of the fiscal 2019 National Defense Authorization Act: Huawei, ZTE, Hytera, Hangzhou Hikvision and Dahua, or their subsidiaries and affiliates. The phrase is a procurement term. It describes who built the hardware, not how well the hardware works or how securely it is configured. It is also a narrower claim than it sounds, because the FCC’s Covered List and Section 889’s list of manufacturers are two different documents and only the second is what the phrase refers to. The Covered List carries those five manufacturers, and also Kaspersky, several international telecommunications carriers, foreign-produced drones and routers, power inverters, and advanced robotic devices. A supplier writing “NDAA compliant” on a quote is answering a question about five named manufacturers. It is not a statement about the Covered List as it stands today. There is a qualifier in the [Covered List](https://www.fcc.gov/supplychain/coveredlist) that almost no product page reproduces. Video surveillance equipment from Hytera, Hikvision and Dahua was added on 12 March 2021, and it sits on the list when used for public safety, security of government facilities, physical security surveillance of critical infrastructure, and other national security purposes. The same wording governs the federal procurement clause at [FAR 52.204-25](https://www.acquisition.gov/far/52.204-25). Telecommunications equipment from Huawei and ZTE carries no such qualifier. The video surveillance entries do. That third purpose has been unsettled for two years. The D.C. Circuit vacated the FCC’s definition of critical infrastructure in 2024, calling it unjustifiably broad, and sent it back to the Commission. With no replacement definition in force, the June 2026 notice suspended the import and marketing prohibition for that one purpose until the FCC adopts one. On 22 July 2026 it did, taking the definition Congress already wrote into the USA PATRIOT Act: systems and assets so vital that their incapacity would have a debilitating impact on security, national economic security, or national public health and safety. The suspension lifts when that definition takes effect. ## Import and marketing is not use Three groups read the same headline and reach different conclusions, and only one of them is under a deadline. > The FCC limited an authorization. It did not switch off a camera. Importers and resellers are directly affected: the covered models cannot lawfully come in or be sold. Federal contractors have a separate, older obligation that has nothing to do with the July date. Everyone else — the private warehouse, the yard, the plant, the distribution centre — has a supply question, not a compliance emergency. When the current stock of covered spares runs out, replacements will have to come from somewhere else. That is a procurement timeline, not a removal order. Instrument What it prohibits Who it binds FCC Covered List (video surveillance entries added 12 Mar 2021) Nothing on its own — it designates equipment and scopes it by purpose of use No one directly Secure Equipment Act rules (from 11 Nov 2022) New FCC equipment authorizations for covered models Manufacturers seeking authorization Public Notice DA 26-635 (effective 16 Jul 2026) Importation and marketing of previously authorised covered models Importers, distributors, resellers Section 889 / FAR 52.204-25 (from 13 Aug 2020) Use of covered equipment as a substantial or essential component of any system Federal agencies and their contractors Covered camera equipment  /  What each instrument actually does Four rules, constantly conflated One reaches use Three do not 13 Aug 2020 Section 889 / FAR 52.204-25 Use of covered equipment as a substantial or essential component of any system Contractors 12 Mar 2021 FCC Covered List Nothing on its own — it designates equipment and scopes it by purpose of use Nobody directly 11 Nov 2022 Secure Equipment Act rules New FCC equipment authorisations for covered models Manufacturers 16 Jul 2026 Public Notice DA 26-635 Importation and marketing of previously authorised covered models Importers Only Section 889 reaches use — and it has been in force since 2020. Sources: FCC Covered List · Public Notice DA‑26‑635 · FAR 52.204‑25 Read the last row carefully, because it is the one that reaches use. Section 889(a)(1)(B) bars an agency from contracting with an entity that uses covered equipment, and that prohibition follows the contractor rather than the contract: the camera on a warehouse wall counts even if no federal work happens in that warehouse. If you hold federal contracts, or intend to bid, this is your obligation and it has been in force since August 2020. ## Why the replace-everything reflex costs more than it fixes The instinct to rip the fleet out is not irrational, and the people acting on it usually have a real prompt. A prime contractor pushed a flow-down clause. An insurer’s questionnaire asked about Chinese-manufactured equipment. A large customer’s vendor security review flagged it. Diligence ahead of a sale raised it. Those are legitimate reasons to replace hardware, and they are all commercial or contractual reasons rather than FCC ones. The problem is what the reflex does to sequencing. A fleet refresh is a capital line, a procurement cycle, and weeks of ladder work across every site. Everything downstream of the cameras waits for it. The analytics project that would have told you something useful about the loading bay this quarter now sits behind a hardware program that finishes next year, and the intervening months produce exactly as much operational insight as the years before them did: none. It also solves less than people expect. Replacing a covered camera with a rebranded unit built by the same original manufacturer changes the label and nothing else. And a fleet refresh chosen purely on brand compliance tends to optimise for the wrong variable, because the specification that decides whether analytics work on a camera is placement and pixel density, not the badge. That is the argument for treating the two decisions separately, which is also [why you don’t need new cameras to get AI video analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) in the first place. ## How to separate the camera decision from the analytics decision Four steps, in this order. The first two are free and usually end the conversation. ### Establish whether the equipment is covered at all Pull the FCC ID from a sample of each model in the fleet. It is printed on the camera label and repeated in the device’s own web interface. That ID identifies the grantee of the equipment authorization, which is the manufacturer of record regardless of whose logo is moulded into the housing. Then apply the purpose qualifier. For video surveillance equipment the Covered List designation attaches to use for public safety, security of government facilities, physical security surveillance of critical infrastructure, and other national security purposes. A camera watching a dock door at a third-party logistics site is a different case from one watching a substation fence. If your operation falls inside one of the sixteen critical infrastructure sectors, treat the qualifier as satisfied and move on; if it plainly does not, record the reasoning and the date, because that record is what you will be asked for later. ### Establish whether you have a use obligation This is a yes or no question with a short decision path. Do you hold a federal prime contract or subcontract? Do you intend to bid for one? Has a customer flowed FAR 52.204-25 down to you in writing? If all three answers are no, you have no federal use obligation, and the July 2026 rule gives you a supply constraint rather than a removal deadline. If any answer is yes, the obligation is real, it predates this year’s news by six years, and the exception and waiver paths are narrow. That is a question for counsel and for your contracting officer, not one to settle from a vendor’s compliance page. ### Deal with the OEM problem before you buy anything A large share of the market is built by a small number of factories. Rebranding is normal and legal, and it is also the reason a compliance-driven refresh can end where it started. Ask for the FCC ID of every proposed model in writing, before purchase, and check it yourself rather than accepting a compliance declaration. Firmware gives a second signal: web interface layout, default port assignments, and the manufacturer string returned by an ONVIF device information query frequently match the original builder rather than the brand on the box. If a supplier will not put the FCC ID in the quote, that is your answer. ### Sequence the refresh so it does not hold the analytics project hostage Analytics that reads a standard RTSP or ONVIF stream does not care which company assembled the camera producing it. That is the practical consequence of the separation, and it runs in both directions. It means an analytics deployment made this quarter survives a hardware refresh made in two years, because the software reads whatever feed exists at the time. It also means the refresh does not have to complete before the analytics start. Run the two as independent programs with independent budgets: replace cameras when the contract, the insurer, or the end of the spares supply requires it, and connect analytics to the fleet you have now. The one dependency worth respecting is the reverse of the usual assumption — knowing which detections you actually want, and at what pixel density, is useful input into the camera specification when the refresh does happen. Working out [what existing cameras can already tell you](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) is therefore a reasonable thing to do first, not last. ## Where Nsightify fits Nsightify reads the IP and CCTV cameras a site already operates. No camera is replaced to connect it, and the analysis is indifferent to the manufacturer, because what it consumes is a standard video stream. Detections such as intrusion into a restricted zone, loitering, blocked exits and forklift–pedestrian proximity produce real-time alerts on the fleet as it stands today, and on whatever fleet replaces it later. Deployment is either Nsightify Cloud or a Zero Trust on-premises option for organisations whose footage cannot leave their own network, which is a common constraint in exactly the sectors where the covered-equipment question arises. To be clear about what this is not: Nsightify does not audit a camera fleet, does not determine whether a given model is covered equipment, and resolves nobody’s Section 889 obligations. Those are questions for your contracting officer and your counsel. The technical limits are worth stating too. Camera-based detection depends on sightlines, lighting, and where a camera was mounted, which was almost never for analytics. Occlusion in a racked aisle or behind a parked trailer is real. Resolution at the distance that matters sets a floor on what can be detected reliably. Those constraints apply to the cameras you own and to any cameras you buy, which is one more reason the [real-time monitoring question](/nsights/physical-safety-monitoring-real-time-ai/) deserves its own answer rather than being folded into a procurement decision about brands. Lighting also changes with the clock: [video analytics at night](/nsights/video-analytics-at-night/) works on an unevenly lit, shorter-range image. ## Questions operators are asking ### Do I have to replace my Hikvision or Dahua cameras? Not because of the July 2026 FCC rule. That rule prohibits importing and marketing previously authorised covered equipment. It does not revoke the authorizations already granted, and the Federal Register notice is explicit that equipment legally purchased and still holding an authorization may continue to be used. A replacement obligation, where one exists, comes from a federal contract clause, an insurer, or a customer requirement, not from the FCC. ### Are my existing cameras still legal to use? For a private commercial operator with no federal contract, yes. Nothing in the June 2026 Public Notice or its Federal Register publication reaches the operation of installed equipment. The Commission limited the scope of existing equipment authorizations under 47 CFR 2.939(e) precisely so it could stop imports without revoking authorizations and forcing equipment off the air. ### What does NDAA compliant mean for cameras? It means the camera contains no equipment produced by the manufacturers named in Section 889 of the fiscal 2019 National Defense Authorization Act: Huawei, ZTE, Hytera, Hangzhou Hikvision and Dahua, or their subsidiaries and affiliates. It is a procurement term describing who made the hardware. It says nothing about image quality, cyber hygiene, or whether the camera suits your site. ### Does Section 889 apply to private companies? Only through federal contracting. Section 889(a)(1)(B) prohibits an executive agency from contracting with an entity that uses covered equipment as a substantial or essential component of any system, and that reaches use outside the contract as well as inside it. A private company with no federal contracts and no intention of bidding for one is not bound by it. ### How do I tell if a camera is an OEM rebrand? The brand on the housing is not evidence. Check the FCC ID printed on the label against the FCC equipment authorization database, which names the grantee rather than the reseller. Firmware fingerprints help too: the web interface layout, default port assignments and ONVIF device information often match the original manufacturer. Ask your integrator for the FCC ID in writing before purchase. * * * ## What to check before your next camera quote Sample the FCC IDs across your fleet and find out what you actually own. Write down whether your operation falls inside the purpose qualifier, and date it. Answer the federal contracting question once, in writing, so it stops being reopened every time a headline lands. If the answer is that nothing forces your hand, you have a spares-supply plan to make over the next few years rather than a capital program to fund this one. Then ask the separate question, the one the hardware debate has been standing in front of: whether anyone is watching the feeds those cameras already produce. For an outside view on what those feeds could tell you, [send the Nsightify team a question](/contact/?interest=question). **Keep reading:** * [You don’t need new cameras to get AI video analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) * [AI video analytics: turn existing cameras into intelligence](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) * [Real-time physical safety monitoring on cameras you own](/nsights/physical-safety-monitoring-real-time-ai/) More on this from Nsightify: [AI video analytics on existing IP and CCTV cameras](/solutions/). [Previous](/nsights/osha-warehouse-emphasis-program-2026/) [OSHA's Warehouse National Emphasis Program Now Runs to 2031](/nsights/osha-warehouse-emphasis-program-2026/) [Next](/nsights/california-ai-workplace-law-sb-947/) [California AI Workplace Law SB 947: A Human Must Corroborate](/nsights/california-ai-workplace-law-sb-947/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## The Nsightify Pilot Program Is Open — Become a Founding Partner URL: https://nsightify.com/nsights/nsightify-pilot-program-founding-partner/ A limited pilot for a small group of operations and security teams — real-time intelligence on your existing cameras. Founding partner spots are limited. [Home](/) [Nsights](/nsights/) The Nsightify Pilot Program Is Open — Become a Founding Partner June 22, 2026 3 min. read # The Nsightify Pilot Program Is Open — Become a Founding Partner News AI Video AnalyticsCompany NewsPhysical SafetyPerimeter Security ![Wide view of a logistics and warehouse facility at dawn with workers walking toward an illuminated loading bay](/_astro/logistics-facility-operations-dawn.7RzeaFh-_Z1FG1YW.webp) We’re opening the doors to a small group of founding pilot partners — and we’re looking for operations and security teams who’d rather shape a platform than wait for one. This is your chance to put real-time intelligence on the cameras you already own, work directly with the team building it, and help define where Nsightify goes next. Spots are limited by design. Here’s what’s on the table. ## What Founding Pilot Partners Get This isn’t a free trial you log into alone and figure out yourself. A pilot partnership means we’re in it with you: * **Real intelligence on your existing cameras — fast.** Connect the IP cameras you already operate (directly, or through a lightweight Nsightify Edge Relay) and start getting [physical safety](/solutions/physical-safety/) and [perimeter security](/solutions/perimeter-security/) alerts. No new hardware. No rip-and-replace. * **Direct access to the team building the platform.** You won’t be routed through a support queue. You’ll work hands-on with the people writing the roadmap. * **A real say in what gets built.** Founding partners shape priorities. The features you need most are the ones we’ll be focused on — because you’re the reason they exist. * **Founding-partner standing.** Get in before everyone else, and carry the advantages of being first as the platform grows. You bring the cameras and the operational reality. We bring the platform and the attention. Together we prove what it looks like when your existing infrastructure finally tells you something useful in real time. ## Who This Is For The pilot is built for teams in asset-intensive environments where what happens on the floor, in the yard, and at the gate actually matters: * Construction sites * Logistics and storage operations * Manufacturing floors * Security and surveillance operations If you’ve already invested in cameras and gotten nothing analytical back from them, you’re exactly who we want to talk to. ## What’s Involved on Your Side We’ve kept the lift deliberately light, because the whole point of Nsightify is that you don’t have to change what you do: * **Connect your existing cameras** — directly to the platform, or via the Nsightify Edge Relay where a direct connection isn’t practical. * **Tell us your priorities** — [physical safety](/solutions/physical-safety/), [perimeter security](/solutions/perimeter-security/), or both. * **Start getting real-time alerts** — no long training period, no data science team required on your end. For most partners, that’s a subscription and the cameras already on the wall. For enterprises and regulated operations with stricter requirements, a Zero Trust deployment keeps everything inside your own private network. ## Why Now — and Why “Limited” Is the Point We’re keeping this group small on purpose. Founding partners get a level of attention that simply doesn’t scale to a crowd — direct access, real influence, and a head start. That only works with a handful of teams, which means the spots are genuinely limited and they’re filling on a first-come basis. If you’ve read how this works and want to see it on your own feeds, this is the moment to raise your hand. ## New to Nsightify? Start Here. If you’re just meeting us, these three pieces explain exactly what the platform does and why it works the way it does — read them, then come back and claim a spot: * [AI video analytics: turn existing cameras into intelligence](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) — what video analytics actually means now. * [Real-time physical safety monitoring on cameras you own](/nsights/physical-safety-monitoring-real-time-ai/) — real-time physical safety on existing cameras. * [You Don’t Need New Cameras to Get AI Video Analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) — why there’s no rip-and-replace. * * * _Founding pilot spots are limited and filling now. [Apply to the Nsightify Pilot Program](/contact/?interest=pilot) and let’s put real-time intelligence on the cameras you already own._ **Keep reading:** * [AI video analytics: turn existing cameras into intelligence](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) * [Real-time physical safety monitoring on cameras you own](/nsights/physical-safety-monitoring-real-time-ai/) * [You Don’t Need New Cameras to Get AI Video Analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) [Previous](/nsights/physical-safety-monitoring-real-time-ai/) [Real-Time Physical Safety Monitoring on Cameras You Own](/nsights/physical-safety-monitoring-real-time-ai/) [Next](/nsights/you-dont-need-new-cameras-ai-video-analytics/) [You Don't Need New Cameras to Get AI Video Analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## OSHA's Warehouse National Emphasis Program Now Runs to 2031 URL: https://nsightify.com/nsights/osha-warehouse-emphasis-program-2026/ OSHA replaced its warehouse national emphasis program in 2026. It runs to 2031, drops retail, and voids old targeting lists. What changed for you. [Home](/) [Nsights](/nsights/) OSHA's Warehouse National Emphasis Program Now Runs to 2031 August 5, 2026 7 min. read # OSHA's Warehouse National Emphasis Program Now Runs to 2031 Physical Safety OSHAWarehouse Operations ![Empty loading dock aisle with a marked pedestrian walkway and an illuminated exit route sign overhead.](/_astro/osha-warehouse-emphasis-program-2026-hero.dIFz7tYk_Z2l4VYK.webp) OSHA’s warehouse national emphasis program lapsed on 13 July 2026 and was replaced eighteen days later by a directive that runs five years instead of three. The replacement is narrower than the program it supersedes, and narrower in ways that change who gets inspected and what an inspector arrives looking for. If your facility was on the old targeting list, it is not automatically on the new one. ## What the 2026 OSHA warehouse national emphasis program replaced A national emphasis program is a temporary OSHA directive that concentrates inspection resources on a specific hazard or industry for a fixed period. It creates no new legal duties. What it sets is which establishments get selected for programmed inspection, and which hazards a compliance officer is instructed to examine once inside. The [2023 instruction](https://www.osha.gov/sites/default/files/enforcement/directives/CPL_03-00-026.pdf) terminated on schedule, three years after its effective date. Its replacement, [CPL-03-00-026](https://www.osha.gov/sites/default/files/enforcement/directives/CPL-03-00-026.pdf), was signed on 6 July 2026, took effect on 31 July 2026, and expires five years from that date — 31 July 2031. It explicitly supersedes the July 2023 version. Worth knowing if you go looking: OSHA’s [directives index](https://www.osha.gov/enforcement/directives/nep) still lists the 2023 directive under its original date, and still links to the old PDF. The superseding document sits at a near-identical filename. Anyone checking the index rather than the document will conclude the program simply expired. OSHA CPL‑03‑00‑026  /  Program terms compared Eighteen days with no emphasis program in effect 2023 term: three years 2026 term: five years 13 Jul 202313 Jul 2026 2023 NEP · three years signed 6 Jul 2026 18 days, no NEP in effect 13 Jul 2026 → 31 Jul 2026 2026 NEP · five years 31 Jul 202631 Jul 2031 2024 2025 2026 2027 2028 2029 2030 2031 The replacement runs two years longer, and starts eighteen days after the old one lapsed. Source: OSHA Instruction CPL‑03‑00‑026 ## Narrower is not the same as lighter The natural reading of a shorter scope is reduced pressure. That reading misses what actually got cut and what got extended. The new directive runs five years rather than three, and OSHA states its case for continuing with its own enforcement record: within the first eighteen months of the 2023 program, the agency reports more than 1,700 violations identified and roughly 37,410 workers removed from hazards. It also cites 2020–2024 five-year average injury rates showing couriers and express delivery carrying an average recordable case rate of 8.6 against 2.6 for all private industry. > The scope narrowed. The clock got longer, and the reasoning got more specific. ## What changed, precisely The directive lists its own significant changes. Five of them matter operationally. Change What it means on the ground High-injury-rate retail removed Home centres, hardware stores, supermarkets and warehouse clubs are out of the program’s scope Mandatory heat and ergonomic screening removed No longer a required step in every inspection under this program Term extended to five years Runs to 31 July 2031, with a program review inside four years Old targeting lists void Area offices need not finish unfinished cycles from the 2023 program OIS code changed to “WAREHOUSE” Prior program inspections stay coded “WAREHOUSE23” ### Retail is out The 2023 program carried a second table of high-injury-rate retail establishments — home centres, hardware stores, other building material dealers, supermarkets, and warehouse clubs and supercenters — subject to partial inspections of storage and loading areas. That table is gone. The new program covers seven NAICS codes: postal processing and distribution centres, couriers and express delivery, local messengers and local delivery, and four warehousing and storage categories. OSHA CPL‑03‑00‑026  /  National Emphasis Program Warehousing & Distribution Centers: 2026 scope Signed 6 July 2026 Effective 31 July 2026 Covered 7 codes · retained * 491110 Postal Service (P&DCs only) * 492110 Couriers and Express Delivery Services * 492210 Local Messengers and Local Delivery * 493110 General Warehousing and Storage * 493120 Refrigerated Warehousing and Storage * 493130 Farm Product Warehousing and Storage * 493190 Other Warehousing and Storage Table 2. Comprehensive safety inspections; the only codes an inspection may be opened under. Removed in 2026 5 codes · retail * 444110 Home Centers * 444130 Hardware Stores * 444190 Other Building Material Dealers * 445110 Supermarkets and Other Grocery (except Convenience) Stores * 452311 Warehouse Clubs and Supercenters Table 3 of the 2023 NEP. High injury rate retail establishments are no longer covered. Retail coverage removed effective 31 July 2026. Source: OSHA Instruction CPL‑03‑00‑026, Tables 2 & 3 If you run a retail operation that was inside the old scope, this program no longer reaches you. Subparts D, E, L and N still do. ### Heat and ergonomic screening is no longer mandatory Under the 2023 directive, heat and ergonomic hazards had to be considered and documented during every inspection, with a health inspection opened where present. The new directive removes that mandatory screening. Heat has not disappeared from the document. It still appears among the hazards inspections focus on, and area offices may expand an inspection to address it. What changed is that the screening step is no longer compulsory on every visit. Heat also remains governed by its own separate directive, which OSHA reissued in April 2026 on its own schedule. ### State plans: read this one carefully The 2023 directive required state plans to participate. The 2026 version is harder to characterise, because the document says two things. Its abstract and the heading of its federal program change section both state that notice of intent and adoption are required. The body of that same section says state plans are “strongly encouraged, but are not required” to adopt the updated instruction, while also setting the 60-day notice and six-month adoption timetable. That is what the instrument says. It is not a settled question, and if your obligations turn on it, it is a question for counsel and for your state plan — not one to resolve from the federal text alone. ### Your old list does not carry over Area offices are not required to complete unfinished cycles from the 2023 program. A new master list gets generated from the seven covered NAICS codes, randomised, and worked through. One deletion rule is worth knowing: establishments that had a comprehensive inspection covering these hazards within the previous three years come off the list. If OSHA inspected you comprehensively in 2024 or 2025, that may keep you off this cycle. It does not affect complaints, referrals, or fatality and catastrophe responses, which reach any covered establishment regardless. ## Where Nsightify fits The hazard focus did not narrow. Powered industrial vehicle operations, material handling and storage, walking-working surfaces, means of egress and fire protection all carry forward. Two of those — egress and vehicle operations — are conditions that either exist or don’t at a given moment on a floor, which makes them observable continuously rather than at audit time. Nsightify’s [Physical Safety](/solutions/physical-safety/) capabilities run on the IP and CCTV cameras already installed, with no camera replacement: blocked-exit detection, forklift-pedestrian proximity and restricted-zone intrusion, with real-time alerting when a condition appears. The constraints are worth stating. Camera-based detection depends on sightlines, lighting, and where a camera was mounted — usually for a purpose other than analytics. Occlusion in a racked aisle is real. And detection is not prevention: an alert says a condition exists now, which matters because the alternative is reading about it in a 301 report later. ## Questions operators are asking ### Is the OSHA warehouse national emphasis program still in effect in 2026? Yes. The 2023 directive terminated on 13 July 2026, and a superseding instruction took effect on 31 July 2026. The new program expires five years from that date, on 31 July 2031. OSHA’s directives index had not yet been updated to reflect the change, so the index alone will give you the wrong answer. ### Does the new program still cover retail establishments? No. The 2023 version included a separate table of high-injury-rate retail establishments — home centres, hardware stores, building material dealers, supermarkets, and warehouse clubs — for partial inspections of storage and loading areas. The 2026 version removes that coverage entirely and applies to seven warehousing, postal and courier NAICS codes. ### Does my state plan have to adopt the updated program? The directive is internally inconsistent on this. Its abstract and section heading state that adoption is required; the body of the same section says state plans are strongly encouraged but not required, while still setting notice and adoption deadlines. Treat the answer as unsettled, check your own state plan’s published response, and take legal advice if your obligations depend on it. ### Is my facility still on OSHA’s warehouse targeting list? Not necessarily. Area offices are not required to finish unfinished cycles from the 2023 program, and a fresh master list is generated under the new one. Establishments that received a comprehensive inspection covering these hazards in the previous three years are removed from the list. Complaints, referrals and fatality responses are unaffected. ### What records can an inspector still ask for? OSHA 300 logs, 300A summaries and 301 incident reports for the current and previous three calendar years, reviewed at the opening conference to identify recorded injuries tied to the covered hazards. That requirement carried over unchanged, and Part 1904 recordkeeping duties run independently of any emphasis program. * * * ## What to check this quarter Confirm your NAICS code against the new table of seven — if you were covered through the retail table, your position changed on 31 July. Check whether a comprehensive inspection in the last three years takes you off the current cycle. Walk your egress routes and loading aisles the way a compliance officer would. Then ask the harder question: whether the only time anyone checks those conditions is when a person happens to be walking past. If that walk leaves you with a question about watching egress routes and loading aisles between inspections, [bring it to the Nsightify team](/contact/?interest=question). **Keep reading:** * [Real-time physical safety monitoring on cameras you own](/nsights/physical-safety-monitoring-real-time-ai/) * [PPE detection and hazard-zone monitoring on existing cameras](/solutions/physical-safety/) * [You don’t need new cameras to get AI video analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) More on this from Nsightify: [PPE detection and hazard-zone monitoring](/solutions/physical-safety/). [Previous](/nsights/you-dont-need-new-cameras-ai-video-analytics/) [You Don't Need New Cameras to Get AI Video Analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) [Next](/nsights/ndaa-compliant-cameras-fcc-import-ban/) [NDAA Compliant Cameras: You May Not Need to Replace Yours](/nsights/ndaa-compliant-cameras-fcc-import-ban/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Real-Time Physical Safety Monitoring on Cameras You Own URL: https://nsightify.com/nsights/physical-safety-monitoring-real-time-ai/ Manual safety checks are point-in-time. Hazards aren't. How AI catches PPE gaps, forklift near-misses and hazard-zone breaches in real time. [Home](/) [Nsights](/nsights/) Real-Time Physical Safety Monitoring on Cameras You Own June 20, 2026 5 min. read # Real-Time Physical Safety Monitoring on Cameras You Own Physical Safety AI Video AnalyticsPPE ComplianceComputer Vision ![Worker wearing a hard hat and high-visibility vest walking through a manufacturing floor with an AI-detected hazard zone highlighted](/_astro/worker-ppe-hard-hat-manufacturing-floor.Da9WuQaA_LfCMH.webp) Here’s how most safety programs work. A supervisor does a walkthrough in the morning. Maybe another in the afternoon. They note what they see, flag what’s wrong, and move on to the other ninety things on their list. The rest of the day? Nobody’s watching. And hazards don’t wait for the walkthrough. A worker steps into a swing radius without a hard hat. A forklift rounds a blind corner toward a pedestrian. A spill sits unnoticed by a high-traffic lane. By the time anyone finds out, it’s an incident report — not a prevented incident. The gap between “we checked this morning” and “what’s happening right now” is where injuries, liability, and failed audits live. Closing that gap is exactly what real-time physical safety monitoring is built to do. ## Why Manual Safety Monitoring Keeps Falling Short It isn’t a discipline problem. It’s a coverage problem. A safety walkthrough is a snapshot. It captures one moment, in one place, through one person’s attention. But a busy site is in constant motion across dozens of zones at once. No supervisor can be everywhere, and no one can hold a hundred camera feeds in their head while also doing their actual job. So three things slip through: * **Point-in-time blindness.** Conditions that are fine at 8 a.m. aren’t fine at 2 p.m. — and the next check is hours away. * **Inconsistent enforcement.** Two supervisors, two different standards for what counts as a violation. Compliance becomes a matter of who’s watching. * **After-the-fact discovery.** The most serious events — a fall, an unconscious worker — are often found late, precisely when minutes matter most. You already have cameras covering most of these areas. The footage exists. It’s just not being watched in a way that helps anyone in the moment. ## What Real-Time Physical Safety Monitoring Software Catches This is where computer vision changes the math. Instead of a person sampling a few moments a day, software watches every feed continuously and recognizes the conditions that lead to incidents — the instant they appear. For [physical safety](/solutions/physical-safety/), that means surfacing things like: * **Protective equipment compliance.** Flag when someone is in an area without the hard hat, vest, or other required gear — every time, not just when a supervisor happens to be looking. * **Hazard-zone enforcement.** Know the moment a person enters a restricted or dangerous area — near heavy equipment, an open edge, or an active work zone. * **Forklift–pedestrian proximity alerts.** Detect when a pedestrian comes dangerously close to an active forklift — before the near-miss becomes a collision. * **Vehicles in pedestrian walkways.** Alert the moment a forklift, truck, or van enters a marked pedestrian-only area. Keeping people and vehicles apart is the most enforceable rule on a floor and the hardest one to police by eye. * **Blocked-exit and egress-path detection.** Catch when an emergency exit or evacuation route is obstructed, so it gets cleared before it matters. * **Person-down detection.** Flag a worker collapsed or lying motionless. It matters most where people work alone — night shifts, plant rooms, and the remote corners of a site where the next round is an hour away. * **Commotion and altercation detection.** Identify sudden, erratic movement that signals a fight or panic, and alert the right people immediately. Each of these becomes a real-time alert sent to the people who can act, rather than a clip discovered after the fact. The walkthrough still happens. It’s just no longer the only thing standing between a hazard and a response. ## What This Changes for the People Responsible For an operations or site manager, the payoff is practical and direct: * **Fewer incidents,** because risky conditions get caught and corrected while they’re still just conditions — not events. * **Consistent enforcement,** because the standard is applied the same way across every zone, every shift, every time. * **Cleaner audits,** because you’re working from continuous monitoring and a real record of conditions, not a binder of twice-daily notes. * **Faster response** to the events that genuinely can’t wait. Customers in asset-intensive industries report that the biggest shift isn’t any single feature — it’s moving from _finding out what happened_ to _being told while it’s happening._ That’s the difference between a safety program that documents incidents and one that prevents them. ## You Don’t Need New Cameras to Do This The objection most teams raise first is hardware. They picture new cameras, new sensors, and a capital project they don’t have the budget or appetite for this year. That’s not how Nsightify works. The platform runs on the cameras you already have. Your existing IP cameras connect directly to the Nsightify platform, or stream up through a lightweight Nsightify Edge Relay where a direct connection isn’t practical. Integrations are designed to be effortless, so the system fits the way your site already runs. There’s no long training period and no data science team to staff. You connect the feeds that were already recording and start getting physical safety alerts from them — on day one. It works the same way across the environments where physical safety matters most: construction sites, logistics and storage operations, and manufacturing floors. Same platform, same approach, applied to whatever your cameras already cover. For organizations with stricter compliance requirements, the platform also offers a Zero Trust deployment that keeps everything inside your own private network — but for most teams, getting started is as simple as a subscription and the cameras already on the wall. ## The Check You Didn’t Make Is the One That Matters Your team can’t watch everything, all the time. They were never supposed to. But your cameras already are — and with real-time physical safety monitoring software, those feeds can finally do something more useful than waiting to be reviewed after someone gets hurt. * * * _Want to see real-time physical safety alerts running on your own camera feeds? [Book a demo](/contact/?interest=physical-safety) — or ask about joining the Nsightify pilot program while spots are open._ **Keep reading:** * [AI video analytics: turn existing cameras into intelligence](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) * [You Don’t Need New Cameras to Get AI Video Analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) More on this from Nsightify: [PPE detection and hazard-zone monitoring](/solutions/physical-safety/). [Previous](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) [AI Video Analytics: Turn Existing Cameras Into Intelligence](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) [Next](/nsights/nsightify-pilot-program-founding-partner/) [The Nsightify Pilot Program Is Open — Become a Founding Partner](/nsights/nsightify-pilot-program-founding-partner/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## How to Calculate Pixels per Foot for Video Analytics URL: https://nsightify.com/nsights/pixels-per-foot-video-analytics/ Pixels per foot decides which existing cameras can run video analytics. The formula, the IEC 62676-4 ladder, makers' own minimums and a survey worksheet. [Home](/) [Nsights](/nsights/) How to Calculate Pixels per Foot for Video Analytics October 1, 2026 13 min. read # How to Calculate Pixels per Foot for Video Analytics AI & Security Camera IntegrationAI Video AnalyticsComputer VisionPerimeter Security ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ZXGEfQ.webp) At 150 feet, a 1080p camera with a 90° field of view delivers 6.4 pixels per foot. A person standing there is 36 pixels tall, 3.3% of the frame, under the 4% Axis sets for its own person detection. Pull the zone in to 75 feet and the same camera clears it, with no new hardware. The check takes two numbers per camera: pixels per foot at the far edge of the zone, and the person’s share of the frame. Both come from a datasheet and a spreadsheet, so a whole camera list can be sorted on paper before anyone schedules a pilot. ## Pixels per foot counts the pixels that land on the target Pixels per foot (px/ft) is the number of horizontal image pixels that cover one foot of the scene at a stated distance from the camera. It equals the image’s horizontal pixel count divided by the width of the scene at that distance. IEC 62676-4, the international application standard for video surveillance, uses pixels per meter. One px/ft is 3.28 px/m. Two lines of arithmetic produce it: 1. **Scene width** = 2 × distance × tan(horizontal field of view ÷ 2) 2. **Pixels per foot** = horizontal pixels ÷ scene width Three consequences follow, and each one surprises people who buy cameras by the megapixel. Double the distance and pixels per foot halves. Narrow the field of view and it rises, at the cost of coverage. Move from 1080p to 4K and the pixel count quadruples, but pixels per foot only doubles, because only the width counts. ## The IEC 62676-4 ladder in pixels per foot The 2025 edition of [IEC 62676-4](https://webstore.iec.ch/en/publication/83425), published on October 9, 2025, anchors operational tasks to pixel density on the target plane. It replaced the four levels of the 2014 edition with seven. Axis Communications publishes the values in [its white paper on the 2025 edition](https://whitepapers.axis.com/en-us/pixel-density-based-on-iec-62676-4-2025), dated September 2026. The pixels-per-foot column below is a straight conversion. Level (IEC 62676-4:2025) Pixels per meter Pixels per foot Overview 20 6.1 Outline 40 12.2 Discern 80 24.4 Perceive 125 38.1 Characterize 250 76.2 Validate 500 152.4 Scrutinize 1500 457.2 The 2014 levels, known as DORI, are still in circulation: Detect at 25 px/m, Observe at 63, Recognize at 125 and Identify at 250, or about 8, 19, 38 and 76 px/ft. Axis’s paper on the 2014 edition says new Axis datasheets still quote DORI distances. Its site-planning tool, AXIS Site Designer, still presets the 2014 levels, and 2025 values have to be entered by hand, according to the 2025 paper. Axis is explicit about the ladder’s scope. The values assume a human operator interpreting the image, and for video analytics, the paper says, “other definitions apply.” Axis adds that light direction, lens quality and compression all move the result, and that meeting the model guarantees nothing. So the ladder gives installers, integrators and IT a shared vocabulary for a view. It doesn’t say what a model will detect in it. The article on [what PPE detection can and cannot verify](/nsights/ppe-detection-what-cameras-cannot-verify/) runs into the same limit for hard hats and vests. ## What camera makers publish for their own analytics Where the standard stops, the camera makers’ own analytics documentation picks up. Four of them publish a minimum size for a person, and no two use the same unit. Maker and document How the minimum is stated Minimum for a person Axis, AXIS Object Analytics manual (updated September 2026) Share of image height 4% minimum, 8% recommended Avigilon, analytics design guide (March 2022) Share of image height; pixels on target at 2 MP 5% on H5A and H5A PTZ cameras, 8–15% on dual-head, modular and corner models; 24–36 px/m (Avigilon gives 8–11 px/ft) Hanwha, Wisenet AI white paper (May 2020) Pixels across the object’s shorter side 15 px minimum, 30 recommended Bosch, IVA capabilities per device (October 2023) Pixels at the analytics’ own processing resolution Objects of 256 square pixels (for example 16 × 16) to detect; a person 96 px tall to read PPE attributes Each figure belongs to one product line and one year, so none of them is “the” analytics requirement. Don’t average them, and don’t convert them into one number. Use the one that matches the system you’re evaluating, in its own unit. Two things change the survey. The first is the resolution the analytics actually processes. A share-of-frame minimum, like the one in the [AXIS Object Analytics manual](https://help.axis.com/en-us/axis-object-analytics), doesn’t move when a stream is scaled down. A pixel-count minimum does. Bosch’s [IVA capabilities paper](https://www.keenfinity-group.com/media/pb/media/products_1/video_systems_1/technical_documantation/whitepaper-iva-capabilitiesperdevice-fw980.pdf) documents analytics that shrink the video before analyzing it, for example to 640 × 360, and PPE analytics that run at up to 1280 × 720. So the survey needs the resolution the analytics works on, which may be neither the camera’s rating nor the recording stream. The second is the gap between finding a person and reading what they wear. Within each maker’s own figures, the gap is wide. Bosch detects an object of 16 × 16 pixels and needs a person 96 pixels tall to read PPE attributes. Hanwha’s 2020 paper asks for 15 pixels across a person to detect them, and at least 112 along the long side, 352 recommended, to read attributes. Detection is cheap in pixels, and attributes are expensive. ## What to collect for each camera before you start One row per camera. These eleven columns are the whole worksheet, lettered as they sit in a spreadsheet: 1. **A · Camera:** the ID, the location and the mounting height. 2. **B · Detection and zone:** what you want from this camera, and which part of its view that covers. 3. **C · Target distance:** feet, to the far edge of the zone. 4. **D · Analyzed width:** pixels, of the frame the analytics processes. 5. **E · Analyzed height:** pixels, of the same frame. 6. **F · Field of view:** horizontal degrees, measured, at the zoom set on site. 7. **G · Pixels per foot:** `=D2/(2*C2*TAN(RADIANS(F2/2)))` 8. **H · Person share of frame:** `=G2*5.58/E2`, formatted as a percentage, for a person 5 ft 7 in (5.58 ft) tall. 9. **I · Minimum to meet:** the criterion of the system you’re evaluating, in its own unit. 10. **J · Night check:** whether infrared is on, and whether the person is still distinct at the far edge. 11. **K · Decision:** keep, re-zone or reassign. Columns A to F are inputs, G and H are formulas, and I to K are the survey. People skip column J. ## Six steps from a camera list to a decision ### Step 1: Fix the target distance for each detection Start from the detection, not the camera. For each camera, mark the farthest point where the detection has to work: the end of the fence run, the far edge of the gate apron, the back row of a parking lot. Record it in column C. The common mistake is measuring to the center of the view. Detection fails at the far edge first, so the far edge is the number that matters. ### Step 2: Use the measured field of view, not the focal length Lens arithmetic gives a field of view from the sensor width and the focal length. Real lenses distort. In [Hanwha’s white paper on lens specifications](https://www.hanwhavision.com/wp-content/uploads/2025/10/White-Paper_Differences-Between-Camera-Lens-Specifications-and-Real-World-Performance-1.pdf), published in October 2025, one 5 MP sensor with a 2.4 mm lens calculates to 95° and measures 123°. Using the calculated figure there would overstate pixels per foot by about 69%, averaged across the frame. Take the measured horizontal field of view from the datasheet for column F. On a varifocal camera, record the zoom set on the pole, not the range printed on the box. ### Step 3: Calculate the scene width and pixels per foot Take the fence-line camera from the worked survey below: a 1920-pixel stream with a measured 90° field of view, 150 feet from the far edge of its zone. The scene there is 300 feet wide. That is 6.4 px/ft, or about 21 px/m, which lands at Overview on the 2025 ladder. The mistake here is using the recording resolution when the analytics reads something smaller. Columns D and E hold the analyzed resolution, so ask for it. ### Step 4: Convert pixels per foot into the person’s share of the frame Most analytics minimums are written in terms of the object, so convert. With square pixels, density is the same vertically. A person 5 ft 7 in (1.7 m) tall, at 6.4 px/ft, is about 36 pixels tall. On a 1080-row frame, that is 3.3% of the image height. Now compare it with column I, in that system’s own unit. Against Axis’s criterion, 3.3% misses the 4% minimum. ### Step 5: Sort each camera into one of three decisions * **Keep:** the target clears the minimum with room to spare, at the far edge. * **Re-zone:** shorten the target distance. Narrow the zone to the part of the view that clears, change the zoom, or move the camera to a closer mount. Each of these is cheaper than a new camera. * **Reassign:** the view supports finding a person but not reading what they wear. Run presence detections there, and give attribute detections to a closer camera. The expensive mistake is replacing a camera whose only problem is where it points. ### Step 6: Check the view after dark The arithmetic assumes a lit, sharp target, and each maker sets its own light floor: * **Axis** recommends at least 50 lux across the whole detection area for day-and-night cameras under artificial light. * **Hanwha**’s 2020 paper suggests a minimum of 300 lux. * **Avigilon**’s guide asks for 8 lux on the target. Pull a recorded clip of the far edge after dark, from the stream the analytics reads, with someone walking the zone if you can arrange it. In column J, record whether infrared was on and whether the person was still distinct at the far edge. A camera that clears the pixel minimum at noon can still fail the light minimum at midnight. ## A worked survey of four cameras at one site Consider a monitoring desk adding analytics to one site it already watches: a fence line, a vehicle gate, a parking lot and a lobby. The desk wants presence detections here, such as intrusion and after-hours entry. The figures are illustrative. Each assumes a person 5 ft 7 in (1.7 m) tall, and is checked against Axis’s person criterion of 4% minimum and 8% recommended. Camera Pixels per foot at the target Person in the frame Decision Fence line: 1080p, 90° measured, 150 ft 6.4 (21 px/m, Overview) 36 px, 3.3% Re-zone to 75 ft: 12.8 px/ft, 72 px, 6.6%. The next pole covers the far half. Vehicle gate: 4 MP (2688 × 1520), 45°, 40 ft 81.1 (266 px/m, Characterize) 453 px, 30% Keep. Read at 640 × 360, it drops to 19.3 px/ft, and the person still fills 30% of the frame. Parking lot: 1080p, 110°, 120 ft 5.6 (18 px/m, below Overview) 31 px, 2.9% Re-zone to 60 ft: 11.2 px/ft, 63 px, 5.8%. Hand the back rows to the next camera. Lobby dome: 1080p, 100°, 25 ft 32.2 (106 px/m, Discern) 180 px, 16.7% Keep. None of the four needs replacing, and two need a shorter zone. The fence line at 75 feet still sits under Axis’s recommended 8%, so it’s a candidate for tighter zoom once a second camera takes the far half. For that fence run, [calibrating loitering detection on a fence line](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/) is the next decision after the pixels. No row lands on **reassign**, because this desk wants presence detections only. A site that also wanted PPE at the gate would test the gate against that system’s attribute minimum, which is a far higher bar than detection. The gate is the instructive row. Scaled down to a 640 × 360 frame, its pixels per foot falls fourfold, while the person’s share of the frame doesn’t move. Under a share-of-frame minimum the scaling costs nothing. Under a pixel-count minimum it can decide the outcome. That is why the worksheet records the analyzed resolution rather than the camera’s rating. ## Surveying cameras for Nsightify’s zone detections Nsightify works with existing IP cameras, with no camera replacement required, and the survey is how that holds camera by camera. The three decisions map directly onto how its detections are set up. Zone detections are placed as a shape on each camera’s view. A camera marked **re-zone** gets its zone drawn over the part of the view that clears the minimum, and the far half goes to the next camera. A camera marked **reassign** runs presence detections, such as [intrusion into a restricted zone or after-hours zone entry](/solutions/perimeter-security/), and leaves attribute detections like PPE compliance to a closer view. The stream Nsightify analyzes is configurable per camera, so the worksheet’s analyzed-resolution columns hold a setting to choose per camera. Views drift after a survey. Camera tampering, one of the ready-to-run detections, alerts when a camera has been repositioned, which is also when its zones need a second look. The survey says whether a view can carry a detection. It can’t say how any model performs on it. That takes [testing video analytics on your own cameras](/nsights/how-accurate-is-ai-video-analytics-testing-standard/), on footage from the far edge, after dark. The worksheet also belongs in [the network requirements packet IT approves](/nsights/video-surveillance-network-requirements-it-approval/), because it names the stream each camera will send. ## Frequently asked questions ### What is 40 pixels per foot resolution? Forty pixels per foot means 40 horizontal image pixels cover each foot of the scene at the target distance, about 131 pixels per meter. On the IEC 62676-4:2025 ladder that sits just above Perceive, at 125 px/m. In the 2014 DORI terms, it clears Recognize. At that density a person 5 ft 7 in tall is about 223 pixels tall. Whether analytics detects them depends on that maker’s own minimum. ### What does DORI mean for a camera? DORI stands for Detect, Observe, Recognize and Identify, the four pixel-density levels in the 2014 edition of IEC 62676-4. They are 25, 63, 125 and 250 pixels per meter, or about 8, 19, 38 and 76 pixels per foot. The 2025 edition replaced them with seven levels, from Overview at 20 px/m to Scrutinize at 1500. Both editions assume a person is watching the image. ### How do you calculate pixels per foot? Divide the horizontal pixel count of the stream by the width of the scene at the target distance. Scene width is two times the distance times the tangent of half the horizontal field of view. A 1920-pixel stream with a 90° field of view covers 300 feet of width at 150 feet, so it delivers 6.4 pixels per foot there. Use the measured field of view from the datasheet. ### Does a 4K camera give four times the pixels per foot of 1080p? No. A 4K stream (3840 × 2160) has four times the pixels of 1080p (1920 × 1080), but only twice as many across its width. With the same lens at the same distance, pixels per foot doubles. Halving the distance to the target does the same thing at no hardware cost, which is why a survey often ends with a camera re-aimed rather than replaced. * * * A finished worksheet is the most useful thing to bring to a demo. It shows which cameras can carry which detections today, which need a shorter zone, and which were never pointed at the problem. [Book an Nsightify demo](/contact/?interest=perimeter-security) and bring that list with you. More on this from Nsightify: [AI video analytics on existing IP and CCTV cameras](/solutions/). [Previous](/nsights/employee-monitoring-laws-by-state-cameras/) [Do Employee Monitoring Laws by State Cover Cameras?](/nsights/employee-monitoring-laws-by-state-cameras/) [Next](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A dome security camera mounted above an empty warehouse loading dock with the dock doors closed and a painted floor line.](/_astro/employee-monitoring-laws-by-state-cameras-hero.DTXNao5P_Z27osnE.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ Physical Safety Regulatory Compliance ](/nsights/employee-monitoring-laws-by-state-cameras/) [Do Employee Monitoring Laws by State Cover Cameras?](/nsights/employee-monitoring-laws-by-state-cameras/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## PPE Detection: What a Camera Verifies, and What It Cannot URL: https://nsightify.com/nsights/ppe-detection-what-cameras-cannot-verify/ PPE detection sees whether a hard hat or vest is present. It cannot see the type, class or fit. Where the boundary sits, and how to design around it. [Home](/) [Nsights](/nsights/) PPE Detection: What a Camera Verifies, and What It Cannot September 22, 2026 13 min. read # PPE Detection: What a Camera Verifies, and What It Cannot Physical Safety PPE ComplianceComputer VisionOSHA ![A worker in a high-visibility vest and hard hat seen from a distance down a warehouse aisle, small in frame and too far away for equipment detail to be legible.](/_astro/ppe-detection-what-cameras-cannot-verify-hero.CzOxMOBq_1Y428F.webp) A hard hat is either on a head or it is not, and a camera answers that question all shift without getting bored. The question a PPE policy actually asks is narrower and harder: whether the right equipment, rated for this hazard, was worn correctly by this person in this area. PPE detection answers the first question well and cannot answer the second at all, and knowing exactly where the line falls is the difference between a useful observation layer and a false sign-off. ## The problem: PPE detection is asked to prove more than it can see A PPE program is a documented chain, not a moment. Under OSHA’s general PPE standard at [29 CFR 1910.132](https://www.osha.gov/laws-regs/regulations/standardnumber/1910/1910.132), the employer has to assess the workplace for hazards, select equipment appropriate to those hazards, communicate the selection to affected employees, and ensure the equipment properly fits each employee. The employer then has to verify that the assessment happened through a written certification identifying the workplace evaluated, the person certifying, and the date. Read that list against a camera. Hazard assessment is a judgment made in advance. Selection is a procurement decision. Fit is a physical property of a specific garment on a specific body. Certification is a document. Exactly one item on the list — whether the selected equipment is actually being worn, here, now — is a visual fact in a scene. The gap matters most where inspection attention is already concentrated, which for distribution and storage sites is the territory covered by [OSHA’s warehouse national emphasis program](/nsights/osha-warehouse-emphasis-program-2026/). PPE detection is the automated observation of whether required protective equipment is present on a person in a camera’s field of view, reported with a timestamp and usually a zone. It answers presence and absence continuously, on every camera, on every shift. It is not an assessment, a selection, a fit check, or a certification. That single sentence is the whole boundary, and almost nothing sold in this category draws it. ## Why the usual approach falls short Buyers are rarely naive about this. They arrive having tried the alternatives, and each one leaves the same gap. **The walk-through.** A supervisor with a clipboard checks a floor. This is the only method on the list that can judge fit and read a garment label, which makes it genuinely irreplaceable — and it covers the few minutes it covers. The sample is small and the observed behave differently from the unobserved. **The toolbox talk and the sign-in sheet.** Excellent evidence that training was delivered. No evidence at all about what happened on the floor afterwards. **The vendor datasheet.** Page one of a search for PPE detection is product pages, and they are structured identically: what the product detects, a list of items, an accuracy percentage, a demo button. Not one of them states a limit. The percentages are the tell — they arrive with no test set, no metric definition and no conditions, which makes them unfalsifiable rather than dishonest. Compare that with how the research literature reports the same task. A 2020 study in _Automation in Construction_ by Nath, Behzadan and Paal built an annotated dataset of roughly 1,500 images and around 4,700 worker instances, and its best approach reached [72.3% mean average precision on real-world images](https://www.sciencedirect.com/science/article/abs/pii/S0926580519308325) at eleven frames per second. Models have moved on since 2020 and that figure is not a verdict on any current product. What it demonstrates is the form of an honest claim: a named dataset, a named metric, a stated operating condition. > The percentages are unfalsifiable rather than dishonest. That is worse, not better. ## What good looks like: split the policy into what a camera can observe and what it cannot The useful design is not “buy detection” or “keep the clipboard.” It is knowing which question goes to which instrument. What the policy asks Can a camera answer it? What answers it instead Is the required item being worn, here, now? Yes — continuously, with a timestamp and a zone Nothing else does this at full coverage Is it the right type and class for this hazard? No Procurement specification and a hazard assessment Does it fit this employee properly? No A physical fit check by a person Is it still serviceable, or cracked and expired? No Scheduled inspection and replacement records Was the hazard assessment performed and certified? No The written certification the standard requires ### What a camera answers well Presence and absence of a large, high-contrast item in a predictable position is close to the ideal computer vision task. A hard hat sits at the top of a person, roughly where a detector expects a head. A high-visibility vest covers a large area of the torso in a color chosen precisely because it is conspicuous. Two further things come free with continuous observation and are easy to undervalue. The first is the timestamp: an observation that carries a time and a zone is a record, and a record is what an incident review needs weeks later. The second is coverage of the hours nobody is walking the floor. Neither requires the camera to make a judgment about the garment. ### Where the answer stops: type, class, fit and condition This is where the vendor pages go quiet, and it is not a small gap. **Head protection has two independent axes, and a camera sees neither.** ANSI/ISEA Z89.1-2014 (R2019), the American National Standard for industrial head protection, sorts helmets by **Type** — Type I tested for impact to the top of the head, Type II also tested for lateral impact — and separately by electrical **Class**: Class G tested at 2,200 volts, Class E tested at 20,000 volts, and Class C offering no electrical protection at all, often because it is vented. A worker in a vented Class C helmet near energized equipment is wearing head protection and is wearing the wrong head protection. A detector reports a hard hat in both cases. The distinction is live rather than theoretical. In a [Safety and Health Information Bulletin published on 22 November 2023](https://www.osha.gov/sites/default/files/publications/SAFETY_HELMET_SHIB.pdf), OSHA set out the limits of the traditional hard hat — protection concentrated on the top of the head, and no chin strap, so it can come off in a slip or trip — and encouraged employers to consider modern safety helmets with side-impact protection and retention. The bulletin is advisory and creates no new legal obligation. But if a site decides to act on it, the change from hard hat to safety helmet is one a camera is poorly placed to confirm, because from an overhead aisle view the two occupy the same space on the same head. **High-visibility apparel is worse, because the class is defined by measurement.** [ANSI/ISEA 107](https://safetyequipment.org/public-review-open-for-proposed-revision-of-ansi-isea-107-standard-for-high-visibility-safety-apparel/) is the governing standard for high-visibility safety apparel, and it sorts garments by Type and performance Class. Those classes turn on the measured area of background and retroreflective material and on laboratory photometric performance — properties established in a test lab and attested on a label, not visible in a frame. A camera can see that something bright and yellow-green is being worn. It cannot see square meters of retroreflective tape, and it cannot read certification. The 2020 edition is the one in effect; ISEA put a sixth revision out for public review with a comment period that closed on 11 May 2026, and no new edition has been published since. **Fit and condition are not inferable.** Fit is an explicit duty in 1910.132(d)(1). A chin strap hanging loose, a suspension adjusted wrong, a shell with a hairline crack, a vest washed until the retroreflective tape has stopped performing — all of these are compliant-looking and non-compliant, and all of them read as “present” to a detector. ### The resolution floor, and the part nobody mentions Even the question a camera answers well has a distance beyond which it stops answering. Detection works from pixels. IEC 62676-4:2025 anchors operational tasks to pixel density on the target plane, and the ladder Axis Communications publishes from that standard runs from 20 px/m for overview, through 40 for outline, 80 to discern, 125 to perceive and 250 to characterize, up to 500 and 1500 px/m for the most demanding tasks. A camera placed to give you an overview of an aisle is, by that ladder, placed to tell you a person is there — not to tell you what is on their head. Two honest caveats matter more than the numbers. First, **those values assume a human operator is interpreting the image**, which Axis states plainly; applying them to an algorithm is an extrapolation, not a specification. Second, **no published standard states a pixel density for PPE detection specifically**, so anyone quoting you one has derived it themselves and should say so. Axis is equally direct that the model is a simplification and that light direction, optical quality and compression all move the result. The practical consequence is unglamorous. The camera that gives you good site overview is usually the wrong camera for PPE, and the fix is a camera position and a lens, decided per zone, not a better model. [Calculating pixels per foot at the far edge of each zone](/nsights/pixels-per-foot-video-analytics/) shows which zones need one. ### Designing the split A program that works treats detection as the continuous layer underneath periodic human verification, and writes down which is which. Put the items whose _presence_ carries the risk on cameras, in the zones where the rule applies, and accept that the output is an observation rather than a finding. Keep type and class on the procurement side, where a specification and a receiving check settle them once for everyone, rather than trying to police them per person per day. Keep fit and condition on a scheduled human inspection, because they are physical properties. Then use the detection record to aim the human effort — a zone generating repeated observations at shift change is telling you where to spend the walk-through you were going to do anyway. ### Five questions to put to a PPE detection vendor Ask for written answers. 1. **What exactly does each detection assert?** “Hard hat present” and “approved head protection in use” are different claims and only one of them is true. 2. **What is the metric, the test set and the conditions?** A percentage without all three is not a measurement. 3. **At what pixel density on target does performance fall off**, and how was that established for this detection rather than borrowed from a human-viewing standard? 4. **How does it behave at night, in rain, backlit, and against a busy background** — measured on footage like ours, not on a curated reel. 5. **What does it do when it cannot tell?** A system that reports uncertainty is more useful than one that guesses and reports a clean answer. ## Where Nsightify fits PPE compliance is one of the detections Nsightify runs from day one on the IP and CCTV cameras a site already operates, with per-zone rules, so a requirement that applies on the dock and not in the office is expressed as a zone rather than as a blanket alert. Each observation carries a time and a zone, which is what turns monitoring into a record an incident review can use later. What the detection asserts is deliberately narrow: whether required equipment appears to be present on a person in view. It does not assert that a helmet is Type II, that a vest meets Class 2, that a garment fits, or that anything has been certified. Those stay with procurement, with the hazard assessment, and with a human inspection — and a vendor telling you otherwise is selling past the capability. The limits are real. Detection depends on camera placement, sightlines, lighting, weather and distance, and a camera mounted for aisle overview will often resolve a person without resolving what is on their head — a mounting decision rather than a software one. Occlusion defeats it outright: a worker behind a rack or facing away may not be assessable at all. Nsightify publishes no accuracy figure for PPE detection or any other detection, because a number measured on someone else’s scene tells you little about yours. The way to find out is [to test it the way the standards test analytics](/nsights/how-accurate-is-ai-video-analytics-testing-standard/), on your own cameras. ## Frequently asked questions ### Can AI detect if someone is wearing a hard hat? Yes, presence and absence of head protection is one of the things camera-based detection does most reliably, because a hard hat is a large, high-contrast object in a predictable position. What detection cannot report is which hard hat. The governing standard sorts head protection by impact type and by electrical class, and those distinctions are not visible to a camera at working distance. ### How accurate is PPE detection? There is no single figure, and any vendor quoting one without a test set, a metric and the conditions it was measured under has given you a claim rather than a measurement. Published peer-reviewed work reports its numbers properly: a 2020 study in Automation in Construction reported 72.3% mean average precision on real-world images using its own annotated dataset. Ask for the same disclosure. ### Can a camera tell what class my vest is? No. ANSI/ISEA 107 sorts high-visibility apparel by Type and by performance Class, which are determined by the measured area of background and retroreflective material and by laboratory photometric testing. A camera sees a garment that looks high-visibility. It cannot measure retroreflective area or confirm certification, so garment class stays a procurement and inspection control. ### Does PPE detection work at night? It depends on the light the camera actually has, not on the time of day. Detection works from the image, so anything degrading the image degrades the result: low light, glare, backlighting, rain and heavy compression. A yard lit to allow safe movement is not necessarily lit well enough to resolve what is on a person’s head at forty meters. Test on your own cameras, at night, before relying on it. ### What camera resolution do I need for PPE detection? No published standard states a pixel density for PPE detection specifically. IEC 62676-4:2025 anchors operational tasks to pixel density on the target plane, ranging from 20 px/m for overview up to 1500 px/m for scrutinizing detail, but those figures assume a human operator is interpreting the image. Treat them as a starting point and confirm on your own cameras at the distances you actually need. * * * ## What to write down this week Take your PPE policy and put a mark against every requirement in it: observable continuously by a camera, or not. The list usually splits about one to four, and the short column is the one worth automating. Then check the cameras covering those zones against the distance question rather than the megapixel count, because a person-sized target and a hard-hat-sized target are not the same problem. The zones where the answer is “too far” are the ones to fix first, and they are the same zones worth raising when you ask [the questions a video analytics vendor cannot bluff](/nsights/video-analytics-rfp-vendor-questions/). If the requirement is about keeping people out of a space rather than what they are wearing, [restricted-area detection around machinery](/nsights/machine-guarding-restricted-zone-optical-detection/) is a different and often easier problem. If you want to see what your own cameras can and cannot resolve before committing to anything, [book a demo with the Nsightify team](/contact/?interest=physical-safety). More on this from Nsightify: [PPE detection and hazard-zone monitoring](/solutions/physical-safety/). [Previous](/nsights/california-emotion-ai-workplace-law-ab-1883/) [California Emotion AI Law: The Definition Is Wider Than the Ban](/nsights/california-emotion-ai-workplace-law-ab-1883/) [Next](/nsights/cargo-theft-statistics-2026/) [Cargo Theft Statistics 2026 Show Fewer, Bigger Thefts](/nsights/cargo-theft-statistics-2026/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Video Analytics at Night Sees a Different Image URL: https://nsightify.com/nsights/video-analytics-at-night/ Video analytics at night works on an unevenly lit, shorter-range image. Why detection drops after dark, and how to check a scene's light before a pilot. [Home](/) [Nsights](/nsights/) Video Analytics at Night Sees a Different Image October 6, 2026 12 min. read # Video Analytics at Night Sees a Different Image AI & Security AI Video AnalyticsPerimeter SecurityCamera IntegrationComputer Vision ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_22vGv4.webp) Perimeter analytics is bought for the hours nobody is on site and usually demonstrated in daylight. Video analytics at night works on a different picture: lit by the camera’s own infrared or by whatever site lighting is left, uneven across the frame, and useful over a shorter distance than the daytime view. The clocks go back on November 1, and that picture arrives an hour earlier on the evening shift. ## What changes when a camera switches to night mode When the light drops below what the sensor needs for a usable frame rate, a day-and-night camera removes its infrared-cut filter, switches to grayscale and, if it has one, turns on its infrared illuminator. Color is gone because infrared leaks through all three of the sensor’s color filters. Axis Communications, in its July 2026 white paper on infrared in surveillance, notes a side effect: some materials take unexpected shades of gray, so a dark jacket can look light. Grayscale on its own is not the problem. The same paper calls well-lit infrared video particularly suitable for analytics. What changes is everything around the color: Daytime view Night view Light source Daylight across the whole frame The camera’s own infrared, or the site lighting still on Evenness Shadows, but light everywhere Bright near the light, dark at the edges Usable distance The field of view As far as the light reaches Focus Set for visible light Can shift under near-infrared Noise and blur Low at short exposures Higher, because low light needs longer exposures or more gain Color cues Available Gone Every row on the right is a reason a detection that works at noon can miss at 2 a.m. ## A detector trained in daylight missed most pedestrians at night Pedestrian-detection research has measured this gap directly. In the [NightOwls study](https://www.robots.ox.ac.uk/~vedaldi/assets/pubs/neumann18nightowls.pdf), presented at the Asian Conference on Computer Vision in 2018, a leading detector trained on daytime street images missed 7.36% of pedestrians on daytime data and 63.99% on night images from a vehicle camera. Trained on night images instead, the same detector missed 17.80%. Training accounts for most of that gap. The rest is the dark. Surveillance footage is harder than a windshield camera’s view. The [NightSurveillance benchmark](https://www.ijcai.org/proceedings/2020/0071.pdf), published at IJCAI 2020, was built from 20 surveillance cameras recording from 5 p.m. to 5 a.m. Trained and tested on that footage, the best detector its authors compared still missed 21.73% of pedestrians. The hard cases were the ones a perimeter cares about: * **Small, distant pedestrians:** the best result missed 38.96%. * **Rain:** 32.61%. * **Blurred frames:** 43.23%. Brightness was not the cure. Misses were lowest for pedestrians of medium brightness, so overexposure near a light hurt as well as darkness at the edge. The small, far pedestrian is also where pixels run out first, which is the arithmetic in [how to calculate pixels per foot for video analytics](/nsights/pixels-per-foot-video-analytics/). These are research detectors on public benchmarks, not any vendor’s product figures. Two conclusions still carry over. A daytime result doesn’t predict a night result, and in NightOwls the gap narrowed most once the detector learned from night images. So night performance has to be measured on its own, on night footage. ## Check a lit scene the way the Defense Department does Where a scene is lit by site lighting (pole lights, wall packs, dock lights), there is a published way to check it before anyone runs a pilot. [UFC 4-021-02](https://www.wbdg.org/FFC/DOD/UFC/ufc_4_021_02_2025.pdf), the Defense Department’s design criteria for electronic security systems, in its edition dated July 28, 2025, has video designers verify that each scene of interest has enough light and specify lighting upgrades where it doesn’t. Faceplate illuminance is the light that actually reaches a camera’s sensor after the scene’s surfaces and the lens have each taken a share. UFC 4-021-02 computes it from the scene’s illuminance, the scene’s reflectivity, and the lens’s transmittance and f-number, then compares the result with the minimum the camera needs at its faceplate. The equation is short: **Faceplate illuminance = scene illuminance × reflectivity × lens transmittance ÷ (4 × f-number²)** * **Scene illuminance** comes from light-meter readings across the scene, in foot-candles. * **Reflectivity** depends on the surface. The UFC’s table puts asphalt at 0.07, grass and trees at 0.2, Portland cement concrete at 0.4 and snow at 0.85. * **Lens transmittance** is typically 0.8. * **Camera minimum and f-number** come from the datasheet. Divide a lux figure by 10.76 to get foot-candles. The equation needs the light the camera requires at its faceplate, as in the UFC’s example; if the datasheet doesn’t say where its minimum was measured, ask the maker. The UFC’s worked example is a camera that needs 0.0007 foot-candle at the faceplate, behind an f/1.6 lens, watching dark aircraft on asphalt under an average of 1.2 foot-candles. That delivers about 0.007 foot-candle at the faceplate, more than nine times what the camera needs. Now point the same camera at a distribution center’s trailer yard, also asphalt. Turned around, the equation gives the lowest reading that still passes: camera minimum × 4 × f-number² ÷ (reflectivity × transmittance), about 0.13 foot-candle here. Say the meter averages 1.2 and reads 0.08 at the far row of trailers. The average passes. The far row puts about 0.0004 foot-candle on the faceplate, under the camera’s 0.0007. The UFC also limits how uneven a scene can be. Design for an average-to-minimum ratio no greater than 4:1 within a scene of interest, it says, and video quality degrades noticeably past 8:1. At an average of 1.2 foot-candles, that asks for at least 0.3 at the darkest point, more than double what the camera alone requires. This yard runs at 15:1. Two cautions keep the check honest. Never compare a minimum-illumination figure across makers, who measure it differently, as Axis notes in its Lightfinder white paper. And a light meter reads visible light, so this check covers scenes lit by site lighting. A scene lit by the camera’s own infrared needs the walk test described below. ## Infrared range ends where the illuminator’s light runs out When a camera lights the scene itself, range depends on the camera’s resolution and how far the illumination reaches, according to Axis. The beam is a trade-off. Too narrow, and it whites out the middle of the scene and leaves the edges dark. Too wide, and it loses reach straight ahead. The UFC gives a sense of scale. It puts a 26-watt LED illuminator’s coverage at about 65 feet with a 120-degree flood and about 310 feet with a 10-degree spot. A wide-angle camera covering a whole yard may light only the first stretch of it at night. > The field of view is the daytime promise. The beam is the night one. One camera  /  What it covers by day, what it lights at night Covered by day, unlit at night 26-watt LED illuminator Top-down view, to scale 65 ft 310 ft 400 ft * Field of view, 120° The yard the camera covers by day * Infrared flood, 120° Lights about 65 ft * Infrared spot, 10° Lights about 310 ft down a narrow line * Site light Check it with a light meter Most of the yard the camera covers by day gets no light at night. Coverage: UFC 4-021-02 (2025) §5-4.2 · Yard depth illustrative Infrared also bounces. Light reflecting off a nearby wall, eave or ceiling can flood the image, and dirt, spider webs, water droplets or snow on a dome reflect it straight back into the lens, according to Axis. Insects are drawn to the light, so an illuminator mounted at the lens draws them into the frame. And the lens may not focus both kinds of light alike: Axis notes that conventional lenses often shift focus between visible and near-infrared light. A camera focused at noon can be soft at 2 a.m. ## Site lighting is designed for work, and the work stops at night Most site lighting exists so people can work, and it is switched and scheduled around them. The UFC notes that after-hours lighting inside buildings may be significantly lower than operating lighting. A yard’s pole lights may run on a timer or photocell schedule set for shift operations, and the overnight yard is where a loaded trailer can sit unwatched, one exposure behind [cargo theft statistics for 2026](/nsights/cargo-theft-statistics-2026/). The light check has to run at the hour a detection matters, with the lights as they actually run. On construction sites the federal rule says it outright. [29 CFR 1926.56](https://www.law.cornell.edu/cfr/text/29/1926.56) sets minimum illumination (5 foot-candles for general construction area lighting, 3 for excavation, accessways and active storage, 10 in construction plants and shops) only while any work is in progress. Nothing in the rule requires light once work stops, which is when a jobsite camera is on duty. _Current as of October 5, 2026: OSHA proposed rescinding 1926.56 and the related 1926.26 on July 1, 2025, and has not issued a final rule._ ## How to tell whether a night zone needs light, a move or a new camera The night walk shows which failure a zone has, and each failure has a different first fix. What the night walk shows Likely cause First fix Misses at the far edge only The light runs out Light the far edge, or pull the zone in White-out near a pole light Glare in the frame Move the fixture out of the view and above the camera A soft image only at night Focus shift under infrared Check and refocus on the night image Flare, or insects crossing the lens Illuminator mounted at the lens Mount a separate illuminator away from the camera Misses on a scene a person can read The detector Measure it on your own night footage Unreadable even when lit The camera A faster lens, or a different camera The glare fix is the UFC’s: keep fixtures out of the camera’s view, and above it. A faster lens is the equation working in your favor, because going from f/1.6 to f/1.2 puts about 1.8 times as much light on the sensor. The most different camera is a thermal one. It senses heat, works in complete darkness and is less affected by rain and fog, according to the UFC, which also lists its high cost and its inability to show facial features or fine detail. When the check fails at a single far corner, the cheapest fix is often a light aimed at that corner. ## Testing video analytics at night before a pilot signs off A night acceptance test belongs in the pilot criteria before the pilot starts. It follows the same logic as [testing video analytics on your own cameras](/nsights/how-accurate-is-ai-video-analytics-testing-standard/), run after dark. 1. **Choose the hardest camera for each zone that matters:** the far edge, the light pole in the frame, the corner past the beam. 2. **Run the light check** on lit scenes at the hour the detection matters, with the lights as they actually run. 3. **Walk the zone after dark.** Walk the far edge, cross the zone, and stand in the brightest and darkest spots. Record what the system flags each time, and check focus on the night image. 4. **Repeat in the weather you get:** rain, fog and the first frost. 5. **Count false alerts per camera per night,** separately from the day: insects, headlights sweeping a fence, infrared bouncing off a nearby wall, rain or spider webs on the dome. 6. **Write the night pass mark down** before the first walk, so the result decides the pilot instead of the other way around. ## Running perimeter detections in the night window Nsightify works with existing IP and CCTV cameras, so the night check decides which of them can carry an after-hours detection. The detections that matter in these hours are ready to run: [after-hours zone entry and intrusion into a restricted zone](/solutions/perimeter-security/), loitering and stopped vehicle. Detection scheduling restricts any detection to a night window, and the customer chooses whether evidence is still captured outside it. Fence-climbing and perimeter pacing are included and switched on at commissioning, once tuned against the site’s own footage. Camera tampering alerts when a camera is moved or its view is blocked, and a separate alert fires when a stream drops. None of it changes the physics. Analytics works with the image the camera produces, and a zone a person can’t make out on a monitor at 2 a.m. is a zone to relight or re-aim before anyone automates it. ## Frequently asked questions ### Do security cameras work at night? Most do, by switching to night mode. The camera removes its infrared-cut filter, records in grayscale and, if it has an illuminator, lights the scene with infrared. The image is usable, but it is lit unevenly and reaches only as far as the light does. Whether analytics works on it depends on the light at the far edge of each zone, and only a test after dark shows that. ### How far can a security camera see at night? As far as its light reaches, which is usually less than its daytime field of view. With infrared, range depends on the camera’s resolution and the illuminator’s beam. The Defense Department’s design criteria cite a 26-watt LED illuminator covering about 65 feet with a 120-degree flood and about 310 feet with a 10-degree spot. With site lighting, the darkest point of the zone sets the limit. Measure your own rather than trusting a datasheet range. ### Why does my camera keep missing people at night? Common causes: the far edge of the zone gets too little light, a light or reflection washes out the image, the lens is focused for daylight, the person is too small in the frame, or the detector learned mostly from daytime images. In the 2018 NightOwls study, a detector trained on daytime street images missed 7.36% of pedestrians by day and 63.99% on night images from a vehicle camera. Trained on night images, it missed 17.80%. ### Do I need lights if I have infrared cameras? Often, yes. Infrared lets a camera see without visible light, at the cost of color and range, and a nearby wall, rain or a spider web on the dome can bounce the light back into the lens. Visible lighting extends usable range and lets you check the scene with a light meter. Where infrared fails a walk test, a light aimed at the far edge is often the cheapest fix. * * * Walk your darkest zone after dark, then [book a perimeter security demo](/contact/?interest=perimeter-security) before the clocks go back on November 1. More on this from Nsightify: [AI video analytics on existing IP and CCTV cameras](/solutions/). [Previous](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) [Next](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) [ ![A dome security camera mounted above an empty warehouse loading dock with the dock doors closed and a painted floor line.](/_astro/employee-monitoring-laws-by-state-cameras-hero.DTXNao5P_Z27osnE.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ Physical Safety Regulatory Compliance ](/nsights/employee-monitoring-laws-by-state-cameras/) [Do Employee Monitoring Laws by State Cover Cameras?](/nsights/employee-monitoring-laws-by-state-cameras/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Video Analytics RFP Questions: 9 Tests Vendors Cannot Bluff URL: https://nsightify.com/nsights/video-analytics-rfp-vendor-questions/ Standard AI camera RFPs produce identical vendor scores. Here are 9 technical questions on RTSP, edge bandwidth, and data retention vendors cannot bluff. [Home](/) [Nsights](/nsights/) Video Analytics RFP Questions: 9 Tests Vendors Cannot Bluff August 25, 2026 14 min. read # Video Analytics RFP Questions: 9 Tests Vendors Cannot Bluff AI & Security AI Video AnalyticsCamera IntegrationComputer Vision ![An enterprise edge computing server appliance mounted in an organized IT equipment rack inside a commercial facility, with neatly routed network cables and soft background view of an operations floor.](/_astro/video-analytics-rfp-vendor-questions-hero.Ki1Pnk1f_Z1enGfa.webp) A standard video analytics request for proposal (RFP) is designed to evaluate features, and that is why it fails. When twenty vendors receive a spreadsheet asking whether their software detects intruders, classifies vehicles, or monitors safety zones, all twenty mark the box compliant. The evaluation scores come back nearly identical, procurement selects a winner on licence price, and four months later the deployment stalls in IT review over network bandwidth, edge server requirements, and video privacy policies. The breakdown does not happen because vendors fabricate their feature lists. It happens because a checklist measures what software can theoretically do in a lab demo rather than how it behaves when connected to fifty existing RTSP streams on an enterprise network. ## The problem with standard video analytics RFPs Most enterprise surveillance RFPs treat computer vision software like an office productivity suite. They ask about user permission tiers, dashboard customisation, single sign-on, and export formats. These operational features matter, and they are also the part of the build every vendor has already solved. The constraints that decide whether a deployment survives are physical and architectural: video stream ingestion, decode overhead, WAN uplink limits, local hardware sizing, and data residency. When an RFP fails to test those, the buyer purchases an architectural assumption that breaks under production load. Procurement Evaluation  /  The Deployment Gap What RFPs score vs. what stalls the rollout Surface checklist focus Architectural bottlenecks What the standard RFP scores Theoretical checklist items · All vendors mark compliant Lab Demo Focus 01 Checklist Feature Checklists Generic compliance tick-boxes 02 Checklist Dashboard UI Widget & layout customisation 03 Checklist User Roles SSO & permission tiers 04 Checklist Export Formats MP4 clips & CSV reports VS PRODUCTION LOAD What actually stalls the rollout Physical network & compute constraints · Production failure modes Critical Bottlenecks 01 Physical Limit RTSP Decode Overhead Multi-stream H.264/H.265 GPU decode 02 Physical Limit WAN Uplink Saturation Continuous cloud streaming choke 03 Physical Limit Edge Compute Sizing Local appliance multi-channel limits 04 Physical Limit Data Sovereignty Retention boundaries & model IP Pilots test isolated feeds; production tests ingestion throughput, appliance sizing, and retention governance simultaneously. Source: Nsightify RFP Framework · Architectural Verification Camera analytics pilots stall at production handoff for a consistent reason: the pilot tested isolated camera feeds, while production tests multi-channel network throughput, appliance sizing, and retention governance all at once. To buy a platform that survives IT scrutiny and runs reliably on day one, the RFP must move from functional wish-lists to verifiable architectural tests. ## Why generic feature checklists fall short A generic RFP question asks: “Does your system support real-time intrusion detection?” Every vendor answers yes. That single question hides four critical operational variables: 1. Does the detection run locally on an edge appliance, or does it require continuous cloud video streaming? 2. Does the model require the primary 4K stream at 30 frames per second, or does it run efficiently on a 720p sub-stream? 3. What happens to site alerting when the facility internet connection experiences packet loss or an outage? 4. Who owns the video frames processed by the model, and does the vendor retain rights to use customer footage for public model training? When those variables are left unaddressed in procurement, the hidden costs emerge after contract execution. An organization budgeted for software licences suddenly discovers it must upgrade network switches, install dedicated fiber uplinks, or replace fifty functional cameras because the vendor software requires proprietary edge firmware. > A software checklist verifies vendor marketing; an architectural test verifies what happens to your network. ## The 9 RFP questions vendors cannot bluff To separate production-ready platforms from brittle prototypes, replace generic feature grids with these nine technical questions. Each question targets a specific operational failure mode and requires an exact architectural response. Question Architectural Focus Red Flag Vendor Response 1\. Stream Ingestion ONVIF / RTSP protocol standards “Requires our proprietary camera firmware” 2\. Stream Tiering Primary vs. sub-stream inference “Must ingest uncompressed 4K at 30 FPS” 3\. Fleet Compatibility Mixed-vendor camera support “Compatible only with certified partner models” 4\. Uplink Consumption WAN bandwidth per 10 streams “Requires 4 Mbps continuous upstream per camera” 5\. Offline Resilience Failover during internet outage “Alerting pauses until cloud connection restores” 6\. Compute Footprint Appliance hardware requirements “Hardware sizing provided after purchase” 7\. Data Sovereignty Video retention and local boundaries “Footage buffers in our cloud for 30 days” 8\. Model Training Customer video data ownership “We retain rights to anonymised training clips” 9\. Network Isolation Zero Trust and VLAN compatibility “Appliance requires open outbound internet access” ### 1\. Ingestion: Which RTSP profiles and codecs do you consume without camera agents? Ask the vendor: _Specify the exact video streaming protocols, codecs (H.264, H.265), and ONVIF profiles your platform ingests directly from standard IP cameras without installing custom software or proprietary agents on the camera._ According to published specifications from [ONVIF](https://www.onvif.org/profiles/), Profile S has served as the baseline for IP video streaming, with Profile T governing advanced video streaming (including H.265 compression and HTTPS transport) and Profile M standardising analytics metadata exchange. A modern video analytics platform should ingest standard RTSP streams over unicast and multicast, supporting both H.264 and H.265 compression across standard ONVIF profiles. If a vendor requires side-loading proprietary code onto the camera or demands a specific camera brand, they are selling a closed hardware stack rather than an open analytics layer. ### 2\. Stream tiering: Does inference run against the primary stream or the sub-stream? Ask the vendor: _Does your computer vision pipeline require the primary high-resolution recording stream, or can models execute against secondary sub-streams (e.g., 720p or 1080p at 10–15 FPS)? What is the minimum required pixel density (pixels per foot) for each detection class?_ To check the answer against your own fleet, [calculate pixels per foot for each camera](/nsights/pixels-per-foot-video-analytics/). Sub-stream inference is the practice of running computer vision detection models on a secondary, lower-resolution camera stream (such as 720p or 1080p) while leaving the primary high-resolution stream untouched for recording and evidentiary archive. Running inference on uncompressed 4K primary streams at 30 frames per second consumes massive GPU decode cycles without improving macroscopic detection accuracy. Detections such as zone intrusion, vehicle dwell, blocked exit paths, and person-down events perform with high statistical reliability on 1080p or 720p sub-streams at 10 to 15 frames per second. If a vendor mandates primary 4K streams for standard operational detections, their software will require three to five times more compute hardware per channel. ### 3\. Mixed fleets: How does the platform handle multi-vendor RTSP streams? Ask the vendor: _How does your platform normalize and ingest concurrent RTSP streams across a heterogeneous camera fleet from multiple manufacturers (e.g., Axis, Hanwha, Bosch, and legacy ONVIF cameras) with varying keyframe intervals and timestamp formats?_ Most commercial facilities operate cameras added across different fiscal years, renovations, and vendor contracts. A production-grade analytics platform must handle variable Group of Pictures (GOP) lengths, dynamic bitrate fluctuations, and minor timestamp drift without dropping channels or crashing the ingestion pipeline. Understanding how to add analytics to the [cameras you already own](/nsights/you-dont-need-new-cameras-ai-video-analytics/) avoids unnecessary capital expenditure. If a vendor states that their software requires uniform camera hardware across the entire facility, the procurement team must factor a full camera replacement into the total cost of ownership. ### 4\. Bandwidth: What is the continuous WAN uplink requirement per 10 camera channels? Ask the vendor: _State the continuous outbound WAN bandwidth consumption required per 10 camera channels during normal operation, and specify whether raw video or structured metadata is transmitted across the external gateway._ Continuous cloud video streaming is bandwidth-intensive. Streaming ten 1080p camera feeds at 4 Mbps consumes 40 Mbps of dedicated upstream bandwidth. Fifty cameras require 200 Mbps of continuous, unthrottled outbound capacity. For facilities with standard commercial internet connections, this saturates the uplink and creates contention with business-critical traffic. An on-premises edge architecture processes video locally inside the facility network. Instead of transmitting raw video across the internet, the edge appliance extracts event telemetry and transmits only structured JSON metadata and occasional compressed alert thumbnails. This reduces WAN uplink utilization from 200 Mbps to less than 1 Mbps for fifty cameras. ### 5\. Offline resilience: What happens to detection and alerting when the WAN drops? Ask the vendor: _Describe the operational behavior of the detection, inference, and alerting engines if the site loses external internet connectivity. Are local alerts dispatched to on-site personnel across the LAN during an outage?_ If a platform relies on cloud-based computer vision inference, an internet service interruption disables all safety and perimeter detection across the property. For high-consequence environments—such as manufacturing plants, logistics yards, and critical infrastructure—this creates an unacceptable operational blind spot. An edge-native platform continues to ingest RTSP feeds, execute inference models, and dispatch local webhook or relay alerts across the local area network even when the external WAN connection is completely severed. When the internet connection restores, the appliance syncs its event logs to the central management console without data loss. ### 6\. Compute footprint: What specific appliance hardware runs 50 concurrent streams? Ask the vendor: _Provide the exact server hardware specifications (CPU socket count, RAM, GPU model, and VRAM capacity) required to execute your full detection suite across 25, 50, and 100 concurrent camera channels._ Vague vendor answers such as “runs on standard commercial servers” lead to budget shortfalls. Efficient computer vision architectures utilise hardware-accelerated video decoding (such as NVIDIA DeepStream or Intel QuickSync) and quantised neural network models (INT8 or FP16 precision) to maximize stream density per GPU. Require the vendor to specify: * Number of camera streams processed per GPU (e.g., streams per NVIDIA RTX or A-series card) * Target inference framerate per stream (e.g., 5 FPS, 10 FPS, or 15 FPS) * Total system power draw and rack unit (RU) space requirements ### 7\. Data sovereignty: Where does raw video reside and when is it purged? Ask the vendor: _Where is raw video footage stored during and after processing? Does the platform store video on vendor-hosted cloud servers, and what is the exact retention and deletion schedule for buffered frames?_ The National Institute of Standards and Technology ([NIST SP 800-145](https://csrc.nist.gov/publications/detail/sp/800-145/final)) outlines clear boundaries for data isolation, multi-tenancy, and consumer control in cloud architectures. In regulated sectors—such as healthcare, defense, logistics, and financial services—transmitting continuous employee and customer surveillance video to third-party cloud infrastructure creates substantial regulatory and compliance exposure. A defensible architecture keeps raw video footage entirely within the customer’s security boundary on site. Video frames are held in volatile memory (RAM) only for the milliseconds required for model inference and are immediately discarded, with only event metadata and authorized review clips written to disk. ### 8\. Model training: Do you use customer video feeds to train shared models? Ask the vendor: _Does your software licence or terms of service grant the vendor any right to collect, store, anonymise, or use customer video footage, images, or metadata to train, fine-tune, or benchmark public, foundation, or shared commercial models?_ Many cloud computer vision vendors include standard clauses in their master service agreements granting them broad rights to utilize customer data for “product improvement” and “algorithm training.” For enterprise buyers, this means proprietary facility layouts, operational workflows, and employee movements become training inputs for the vendor’s general product line. Require a contractual guarantee that customer video streams remain private, that zero customer footage is ingested into shared training corpuses, and that all models deployed on your site are pre-trained or fine-tuned strictly within your dedicated tenant boundary. ### 9\. Network isolation: Can the appliance run inside a restricted Zero Trust VLAN? Ask the vendor: _Can the on-premises inference appliance operate within an isolated, non-routable camera VLAN with restricted outbound firewall rules? Provide the complete list of required outbound destination ports, protocols, and fully qualified domain names (FQDNs)._ Enterprise IT security teams enforce strict network segmentation. Security cameras should reside on an isolated VLAN with no direct inbound or outbound internet access. An analytics appliance must sit across the camera VLAN (to ingest RTSP feeds) and the corporate network (to dispatch alerts), while requiring only minimal, encrypted outbound HTTPS (TCP 443) connections for management and telemetry. If a vendor requires opening inbound firewall ports, disabling network firewalls, or routing unsegmented camera traffic directly to public IP addresses, the solution violates standard enterprise security posture. ## How to run an adversarial pilot on existing cameras Do not evaluate video analytics on a single camera pointed at an empty hallway in perfect daylight. An adversarial pilot tests software against the edge cases that cause production systems to generate false alarms or miss critical events. Select five challenging existing camera locations: 1. **High-contrast doorway**: Tests how the model handles harsh backlighting, lens glare, and rapid exposure changes. 2. **Busy loading dock aisle**: Tests object classification and separation when forklifts, pallet jacks, and pedestrians cross in close proximity. 3. **Low-light perimeter fence**: Tests detection sensitivity, infrared illumination noise, and insect motion filtering. 4. **Obstructed machine zone**: Tests zone intrusion accuracy when safety fencing or structural columns partially occlude the field of view. 5. **Weather-exposed yard camera**: Tests false alarm suppression during heavy rain, wind-blown vegetation, and headlights. During the pilot, track three objective metrics: **true positive detection rate**, **false positive alert frequency per 24-hour period**, and **edge compute utilization under peak movement**. A platform that maintains clean detection with less than one false alert per camera per day in challenging conditions is ready for enterprise rollout. ## Where Nsightify fits Nsightify provides an open AI video analytics platform designed to add operational visibility to the IP and CCTV cameras organizations already operate. Our architecture connects directly to existing camera infrastructure over standard RTSP and ONVIF streams, requiring zero camera replacements or proprietary edge firmware modifications. The platform delivers allowlisted, production-ready detections across two core operational solutions: * **Physical Safety**: Ready-to-run detections for PPE compliance, unauthorized vehicles in pedestrian zones, blocked exits and obstructed egress paths, camera tampering, forklift–pedestrian proximity, person down, crowd occupancy limits, restricted-area dwell, and entry into restricted machine areas. * **Perimeter Security**: Ready-to-run detections for zone intrusion, directional tripwires, loitering, after-hours zone entry, camera tampering, tailgating and piggybacking detection, stopped vehicles, and people and vehicle classification. Advanced behavioral detections like perimeter pacing and fence-climbing are included and calibrated during site commissioning. Nsightify deploys through Nsightify Cloud or as a Zero Trust on-premises installation for enterprise and regulated environments. In on-premises deployments, all computer vision inference executes locally on dedicated edge appliances. Raw video never leaves your secure network boundary; only structured event metadata and real-time alerts are transmitted. The technical limits must be stated plainly. Camera analytics depends on optical sightlines, lighting conditions, and mounting angles. Occlusion in tightly packed racking aisles is a physical reality, and detection software acts as an operational monitoring layer rather than a certified safety interlock or automated access enforcement tool. Reviewing [how AI video analytics turns existing cameras into intelligence](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) helps establish practical operational baselines for your facility. ## Questions IT and security teams are asking ### Why do standard video analytics RFPs fail during deployment? Standard RFPs rely on feature checklists where every vendor marks compliant. They fail to test mechanical limits: continuous WAN bandwidth consumption, sub-stream versus primary-stream inference, compute appliance sizing for multi-camera workloads, and customer video data retention boundaries. ### Can video analytics run on sub-streams without losing detection accuracy? Yes. For macroscopic detections like zone intrusion, vehicle counting, and blocked exit paths, 720p or 1080p secondary sub-streams at 10 to 15 frames per second provide sufficient pixel density while reducing compute overhead by over 70 percent compared to 4K primary streams. ### How much uplink bandwidth does cloud video analytics require? Streaming 50 continuous 1080p camera feeds to a cloud platform requires between 150 and 250 Mbps of dedicated outbound WAN bandwidth. On-premises edge inference processes video locally on site and transmits only lightweight JSON event metadata, requiring under 1 Mbps. ### Does adding video analytics require replacing existing IP cameras? No. A vendor with standard ingestion architecture connects directly to existing IP and CCTV cameras via RTSP or ONVIF Profile S and Profile T feeds. If a vendor mandates proprietary camera hardware, the constraint lies in their software packaging rather than an optical requirement. * * * ## What to put in your procurement checklist this week Take your current RFP template and strike out the generic capability tables. Replace them with the nine architectural questions above, requiring vendors to provide specific bandwidth calculations, hardware sizing models, and data governance commitments in writing. Before signing a production contract, insist on a 14-day proof of concept on five of your most challenging existing cameras. Verify that the appliance runs inside your isolated camera VLAN, consumes predictable local compute, and delivers actionable alerts without flooding your security team with false alarms. The Nsightify pilot runs on your own cameras, so it can serve as that proof of concept. [Apply to the pilot program](/contact/?interest=pilot). **Keep reading:** * [You don’t need new cameras to get AI video analytics](/nsights/you-dont-need-new-cameras-ai-video-analytics/) * [How AI video analytics turns existing cameras into intelligence](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) * [NDAA compliant cameras and the FCC covered list: what IT needs to know](/nsights/ndaa-compliant-cameras-fcc-import-ban/) * [AI video analytics on existing IP and CCTV cameras](/solutions/) More on this from Nsightify: [AI video analytics on existing IP and CCTV cameras](/solutions/). [Previous](/nsights/california-ai-workplace-law-sb-947/) [California AI Workplace Law SB 947: A Human Must Corroborate](/nsights/california-ai-workplace-law-sb-947/) [Next](/nsights/machine-guarding-restricted-zone-optical-detection/) [Machine Guarding Restricted Area: Where Optical Detection Fits](/nsights/machine-guarding-restricted-zone-optical-detection/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Video Surveillance Network Requirements: Getting IT to Yes URL: https://nsightify.com/nsights/video-surveillance-network-requirements-it-approval/ Video surveillance network requirements decide whether an analytics rollout ships or stalls: the streams, ports, credentials and egress paths IT approves. [Home](/) [Nsights](/nsights/) Video Surveillance Network Requirements: Getting IT to Yes September 7, 2026 13 min. read # Video Surveillance Network Requirements: Getting IT to Yes AI & Security Camera IntegrationAI Video AnalyticsComputer Vision ![A network equipment room in an industrial facility, with a wall-mounted patch panel of dressed ethernet cabling and an annotated printed floor plan open on a table in the foreground.](/_astro/video-surveillance-network-requirements-it-approval-hero.DOujAaUC_Z1J55R.webp) The budget is signed, the vendor is chosen, and the rollout has a start date. Six weeks later it is still sitting in IT review, because nobody has written down what IT is actually being asked to approve. Video surveillance network requirements are the difference between a two-week review and a two-quarter one, and they are almost never in the proposal. ## The problem: IT is handed a product, not a change The request that reaches the security review queue usually says “approve the video analytics platform.” That is not a reviewable object. A network change approval needs specifics: which streams leave which subnet, to which destination, over which protocol and port, authenticated with which credential, and what happens to each of those when it fails. The reviewer’s caution is not obstruction. It is the documented failure pattern. In their joint advisory [Top Ten Cybersecurity Misconfigurations](https://www.cisa.gov/news-events/cybersecurity-advisories/aa23-278a) (AA23-278A, October 2023), the NSA and CISA rank default configurations first, lack of network segmentation fourth and poor credential hygiene ninth — and they name security cameras among the devices that commonly carry default credentials an attacker can use to move laterally into the rest of the estate. So the person holding up the project is not being difficult about a safety initiative. They are being asked to attach several dozen embedded devices, many of them years past their last firmware review, to a new outbound data path — on the strength of a datasheet. The vocabulary for that decision already exists. The [NIST Cybersecurity Framework 2.0](https://www.nist.gov/cyberframework) (NIST CSWP 29, February 2024) added Govern as a sixth Function alongside Identify, Protect, Detect, Respond and Recover, and placed supplier risk under it as the GV.SC category. The reviewer is being asked for a GV.SC-shaped decision and handed none of the GV.SC-shaped inputs. > IT is not approving software. It is approving a set of connections that will outlive the project team. ## Why the usual approach falls short Three things are usually offered in place of a review packet. Each is reasonable, and none of them answers the question. ### The datasheet answer: “works with any ONVIF camera” ONVIF is a family of profiles, not a single badge. Profile S and Profile T cover video streaming, Profile G covers edge recording, Profile M covers analytics metadata and events, and Profile V — currently a release candidate — covers cloud and video-surveillance-as-a-service deployments. Other profiles in the family govern door and credential systems and have no place in a video analytics review at all. Conformance to one profile guarantees nothing about any other. The badge is also moving underneath the inventory. ONVIF announced in October 2025 that it will [end support for Profile S](https://www.onvif.org/pressrelease/onvif-to-end-support-for-profile-s/), with **31 March 2027** the last date manufacturers can submit products for Profile S conformance. The reason is authentication: Profile S mandates username token authentication, which ONVIF now describes as too weak to protect devices against unauthorised access, and it points implementers at Profile T’s digest authentication or at TLS. Existing deployments keep working — but a camera listed as “ONVIF compliant” on a spreadsheet may be conformant to a profile being retired for a security reason, which is precisely the sort of detail a reviewer is entitled to know before signing. ### The pilot that proved nothing about the network One camera, one laptop, one afternoon, everything on the same VLAN. That pilot demonstrates detection quality, which was never the reviewer’s objection. It answers no question about segmentation, credential handling, egress volume or behaviour during an outage, because the topology it ran on is not the topology it will be deployed on. ### The vendor questionnaire, which tests the vendor Sending the vendor a structured technical questionnaire is a genuine improvement, and the [video analytics RFP questions vendors cannot bluff](/nsights/video-analytics-rfp-vendor-questions/) are the right ones to send. But a questionnaire tests the vendor’s architecture. The approval is a change to your own estate, and the artifacts that justify it are yours to produce — the vendor cannot write your firewall rules or name your time source. ## What good looks like: a video surveillance network requirements packet A video analytics network review is the assessment that decides whether an analytics platform may connect to an existing camera estate. It covers which streams are consumed, over which protocol and ports, using which credentials, across which network boundary, and what data crosses the site perimeter. Assembled as six artifacts, it becomes a two-week review rather than a standing agenda item. ### 1\. A stream inventory, not a camera count Inference consumes streams, not cameras. A single camera commonly publishes a primary stream for recording and a lower-resolution sub-stream, and the analytics platform may take either. “Sixty cameras” is therefore not a specification; “sixty sub-streams at 1080p, H.264, two-second keyframe interval, audio disabled” is. List each stream with its source subnet, codec, resolution, frame rate and the URL form used to reach it. Uplink and compute sizing follow from that list rather than from the camera count, and those calculations belong in the vendor’s response, not in the reviewer’s head. Add each camera’s field of view and distance to the zone, and the same list tells you [which cameras can support a detection, in pixels per foot](/nsights/pixels-per-foot-video-analytics/). ### 2\. Two topologies, and the question that separates them Almost every deployment is one of two shapes: cameras reach the analytics platform directly across the network boundary, or a bridge component sits inside the network, consumes the streams locally and sends outward on their behalf. The distinction is not a product tier. It decides what the firewall change looks like. Review criterion Cameras reach the platform directly Bridge inside the network Firewall change Camera VLAN needs a route outward, or inbound reachability Outbound only, from one known host Credential exposure Camera credentials used across the boundary Camera credentials stay inside the VLAN Rule specificity One rule per camera subnet One rule, one source address If the analytics host is compromised Blast radius is whatever the cameras can reach Blast radius is one host you already monitor During a WAN outage Ingest and alerting stop Local ingest continues, delivery queues Protocol detail decides whether either shape works. RTSP, specified in [IETF RFC 2326](https://www.rfc-editor.org/rfc/rfc2326) (April 1998), uses port 554 by default for the control connection and then negotiates transport for the media: separate UDP ports, or the media interleaved into the same TCP connection. A firewall rule that permits the control connection but not the media path yields a session that sets up cleanly and never delivers a frame — an outcome that reliably gets diagnosed as a broken product rather than a missing rule. RFC 7826 defines RTSP 2.0 and formally obsoletes RFC 2326, but it is not backwards compatible beyond version negotiation, so deployed cameras and the platforms that read them still speak the 1998 protocol. The standards work is moving toward the bridge shape. ONVIF’s Profile V uses its Uplink specification, in which “the conformant device initiates an outbound connection to the cloud-based conformant client over WSS” — a device-initiated, TLS-secured channel that reaches a client through a local firewall without port forwarding. Profile V is a release candidate, so it belongs in a review as a direction of travel rather than as a requirement you can put in a purchase order today. ### 3\. Credentials that survive a hardening cycle The default answer in the field is one shared administrative account, configured at commissioning, used by every integration, and rotated never. That is the ninth item on the NSA and CISA list, and it is how a camera compromise becomes a network compromise. The packet should name a dedicated service account per platform, scoped to stream access rather than device administration, with a rotation procedure that has been tested against the ingest path. Rotation is where this fails quietly: credentials change, streams stop, and nothing tells anyone unless ingest health is monitored as a first-class signal rather than inferred from the absence of alerts. The ONVIF deprecation makes the same point from the other direction. If a manufacturer removes username token authentication in a future firmware release — exactly the hardening a security team would ask for — a device stops being Profile S conformant. A compatibility claim can be invalidated by a security update. Ask which authentication method the platform uses, and whether it survives the camera vendor’s own hardening plans. ### 4\. A time source, decided before it matters Time synchronisation is the field nobody reviews and everybody needs later. Cameras drift. When three cameras covering one loading bay disagree by forty seconds, the event sequence assembled from them is wrong, and the reconstruction of an incident argues with itself. Name the NTP source in the packet, confirm every camera and the analytics host use it, and decide whether the camera VLAN may reach it directly or through an internal server. It is a one-line decision at approval and an expensive forensic problem afterwards. ### 5\. An egress claim you can verify yourself “Only metadata leaves the site” is a testable claim, not a matter of trust, and a reviewer who cannot test it should not be asked to accept it. Four checks settle it: * A deny-by-default egress rule with an explicit allowlist of destinations, each named by fully qualified domain name and port. * Byte counters on that rule, sampled over a normal operating week and compared against the vendor’s stated figure. * DNS query logs from the analytics host, showing every destination it actually resolves. * A packet capture on the uplink over a defined window, confirming the payload shape. The artifact these produce is a data-flow diagram naming every destination. That diagram, not a sentence in a proposal, is what closes the data-residency question — and it is the one document that stays useful for the life of the deployment. ### 6\. Failure modes to name before you sign A review is credible when it has imagined the system breaking. These are the eight worth writing down, because each has a cheap answer at approval and an expensive one afterwards: 1. **A camera reboots and its DHCP lease changes.** The configured stream URL now points at nothing. Fixed with static reservations or name-based addressing, agreed before deployment. 2. **Credentials rotate and ingest stops silently.** No alert fires, because the system’s job is to alert on events and it is now receiving none. 3. **Time drifts and events land out of order.** Correlation across cameras degrades before anyone notices. 4. **A TLS certificate expires** on the management or transport path, and streams fail closed overnight. 5. **A firmware update changes a stream path or removes an authentication method.** The compatibility matrix silently becomes wrong. 6. **Facilities replaces a camera like-for-like** without telling IT, and the asset inventory the approval rests on drifts out of date. 7. **The WAN drops.** Decide in advance what continues locally and what queues, then confirm it rather than assuming it. 8. **The contract ends.** Name which accounts get disabled, which firewall rules get removed and which data gets deleted, at approval — decommissioning rules are the ones nobody ever gets around to removing. ## Where Nsightify fits Nsightify runs AI video analytics against the IP and CCTV cameras a site already operates. The platform ingests standard RTSP streams from existing cameras, so no camera replacement, proprietary firmware or new field cabling is required, and real-time alerts reach the people already watching the site. Both topologies above are supported deliberately. Nsightify Cloud suits sites where an outbound path from a single host is the simplest change to justify. Zero Trust on-premises deployment suits enterprise and regulated environments that require inference to run inside the boundary, with raw video staying on the local network and only structured alert metadata leaving it. Which one a site uses is a network decision made during this review, not a product tier chosen in procurement. The limits belong in the same paragraph as the claims. Camera-based analytics depends on sightlines, mounting height, lens coverage and lighting; occlusion and weather affect what any optical system can resolve. It is an alerting layer that sharpens human response, not a physical barrier or an interlock. Reviewing [how AI video analytics turns existing cameras into intelligence](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) is a reasonable way to set expectations before the network conversation starts. ## Questions IT teams are asking ### Do my cameras need to be ONVIF compliant to add video analytics? Usually not. ONVIF conformance helps with discovery and configuration, but most analytics platforms ingest video over RTSP, which cameras support independently of any ONVIF profile. Check the profile rather than the badge: conformance to one profile implies nothing about another, and ONVIF has set 31 March 2027 as the last date for Profile S conformance submissions because its username token authentication is no longer considered strong enough. ### What ports does RTSP use on a camera network? RTSP is specified in IETF RFC 2326 and uses port 554 by default for the control connection. Media then travels either over separate UDP ports or interleaved into the same TCP connection. That choice matters at a firewall: a rule permitting the control connection but not the media path produces a session that negotiates successfully and delivers no video. ### Should security cameras be on their own VLAN? Yes, and it is the change with the widest effect. NSA and CISA list lack of network segmentation among the ten most common cybersecurity misconfigurations, and name security cameras among the devices that commonly ship with default credentials an attacker can use to move laterally. An isolated camera VLAN limits what a compromised camera can reach and forces the analytics path to be stated explicitly. ### Can I add AI to my existing CCTV system without new hardware? Usually yes. A platform that ingests standard RTSP streams works against the cameras already installed, so the constraint is rarely the camera itself. It is the network path: whether streams can reach the platform, which credentials they use, and what data leaves the site. Those questions decide the rollout, and they are answered on a network diagram rather than a datasheet. * * * ## What to put in your change request this week Before the next review meeting, produce three pages. The first is the stream inventory: every stream the platform will consume, with its source subnet, codec, resolution and frame rate. The second is the data-flow diagram: the topology you are proposing, every destination outside the boundary named by fully qualified domain name and port, and the service account each connection uses. The third is the failure-mode list above, annotated with your answer to each. Those three pages convert an open-ended security objection into a set of decisions with owners. They also outlive the project — the same documents answer the auditor’s question next year and the incident reviewer’s question the year after. If your camera estate is already documented, the [existing-camera analytics question](/nsights/you-dont-need-new-cameras-ai-video-analytics/) is mostly settled and only the network path remains. To test that path on your own topology rather than a datasheet, [apply for an Nsightify pilot](/contact/?interest=pilot). **Keep reading:** * [NDAA compliant cameras and the FCC covered list](/nsights/ndaa-compliant-cameras-fcc-import-ban/) * [Loitering detection false alarms: calibrating dwell time](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/) More on this from Nsightify: [AI video analytics on existing IP and CCTV cameras](/solutions/). [Previous](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/) [Loitering Detection False Alarms: Calibrating Dwell Time](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/) [Next](/nsights/california-workplace-surveillance-camera-placement/) [California Workplace Surveillance Laws: Where Cameras Cannot Go](/nsights/california-workplace-surveillance-camera-placement/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## You Don't Need New Cameras to Get AI Video Analytics URL: https://nsightify.com/nsights/you-dont-need-new-cameras-ai-video-analytics/ Rip-and-replace is why most video analytics projects stall. How to add AI CCTV analytics to existing IP cameras — SaaS-first, Zero Trust optional. [Home](/) [Nsights](/nsights/) You Don't Need New Cameras to Get AI Video Analytics June 22, 2026 4 min. read # You Don't Need New Cameras to Get AI Video Analytics AI & Security AI Video AnalyticsCamera IntegrationPerimeter SecurityPhysical Safety ![Close-up of an existing IP security camera mounted on a concrete wall at an industrial facility](/_astro/existing-ip-camera-closeup.DYAOVL39_ZGeEu4.webp) Every IT and infrastructure leader has heard this pitch before: “Just add intelligent video analytics.” Then the fine print arrives. It needs proprietary cameras. A rack of dedicated hardware. A new vendor relationship. A network redesign. A deployment timeline measured in quarters. So the project goes on the someday list, and the cameras keep recording footage nobody analyzes. The instinct to say no is correct — for that model. Rip-and-replace is genuinely a bad deal. But it’s also no longer the only option. You can add AI-powered CCTV analytics to the cameras you already operate without touching the hardware, and that distinction is the whole story. ## Why Rip-and-Replace Kills Video Analytics Projects The hardware-first model fails for reasons that have nothing to do with whether the analytics are any good: * **Capital cost.** Replacing or augmenting a fleet of working cameras is a budget line that’s hard to justify when the existing ones already do their job. * **Deployment risk.** New hardware on the network means new failure points, new things to secure, and new variables in an environment you’ve worked hard to keep stable. * **Downtime.** Swapping infrastructure on a live site — a working floor, an active perimeter — is disruptive in exactly the places where disruption is most expensive. * **Vendor lock-in.** Proprietary cameras tie you to one supplier’s roadmap, pricing, and pace forever. None of these are analytics problems. They’re infrastructure problems the old model creates. Remove the hardware requirement and most of the reasons to say no disappear with it. ## A SaaS-First Approach to AI-Powered CCTV Analytics Nsightify is built SaaS-first, which means the intelligence lives in the platform — not in a box you have to install, power, secure, and maintain on every site. Connecting your cameras is deliberately simple, and you have two clean paths: * **Direct connection.** Your existing IP cameras connect straight to the Nsightify platform. No proprietary hardware in between. * **Nsightify Edge Relay.** Where a direct connection isn’t practical, a lightweight relay streams your existing camera feeds up to the platform. It’s a thin connector for getting feeds where they need to go — not a heavy on-site appliance you have to manage. Either way, the analysis runs in Nsightify. Integrations are designed to be effortless, so adding analytics doesn’t turn into a standing item on your roadmap. From an infrastructure standpoint, you’re adding a software service that reads feeds you already produce — not a new hardware layer to own. The result is the thing the old model could never offer: real time-to-value. No procurement cycle for cameras, no staging, no multi-month rollout. You connect existing feeds and the platform starts turning them into [physical safety](/solutions/physical-safety/) and [perimeter security](/solutions/perimeter-security/) intelligence. ## When You Need More Control: Zero Trust Deployment SaaS is the right default for most organizations. But some environments — regulated industries, security-sensitive operations, anyone with strict data sovereignty requirements — can’t send feeds to a managed cloud, full stop. That’s a legitimate constraint, not a preference to argue with. For those cases, Nsightify offers a Zero Trust deployment. The same AI-powered CCTV analytics run inside your own private network or environment, keeping data where your compliance posture requires it to stay. You get the intelligence without the data ever leaving your control. The point is that the deployment model bends to your requirements. SaaS-first for simplicity and speed; Zero Trust where the rules demand it. You choose based on your infrastructure and risk posture — not based on what the vendor’s architecture forces on you. ## What This Looks Like on Your Network For the person responsible for keeping the environment clean and secure, the practical picture is straightforward: * **No new camera fleet** to procure, mount, or depreciate. * **No heavy on-site appliance** to patch and maintain across locations. * **A connection model that fits your reality** — direct where it works, an Edge Relay where it doesn’t. * **A deployment choice that fits your compliance needs** — managed SaaS or Zero Trust. You’re not redesigning the network to bolt on analytics. You’re pointing existing feeds at a platform and getting real-time intelligence back. That’s a far smaller surface to evaluate, secure, and sign off on — which is also why these projects actually ship instead of stalling in review. ## The Best Camera Upgrade Is No New Cameras at All The cameras already work. They’re already mounted, already wired, already producing exactly the feeds that AI-powered CCTV analytics needs. The only thing missing is something reading them in real time and telling you what matters. That’s a software problem now — not a hardware project. And solving it shouldn’t require you to rebuild anything you’ve already got running. * * * _Curious how this fits your existing setup? [Book a demo](/contact/?interest=physical-safety) and we’ll walk through connecting your current cameras — direct or via the Edge Relay — and which deployment model fits your environment._ **Keep reading:** * [AI video analytics: turn existing cameras into intelligence](/nsights/ai-video-analytics-turn-cameras-into-intelligence/) * [Real-time physical safety monitoring on cameras you own](/nsights/physical-safety-monitoring-real-time-ai/) More on this from Nsightify: [AI video analytics on existing IP and CCTV cameras](/solutions/). [Previous](/nsights/nsightify-pilot-program-founding-partner/) [The Nsightify Pilot Program Is Open — Become a Founding Partner](/nsights/nsightify-pilot-program-founding-partner/) [Next](/nsights/osha-warehouse-emphasis-program-2026/) [OSHA's Warehouse National Emphasis Program Now Runs to 2031](/nsights/osha-warehouse-emphasis-program-2026/) ## More on topic [All Nsights](/nsights/) [ ![An open logbook and a two-way radio under a desk lamp in a guardhouse at night, with a closed gate and a floodlit fence line in the window.](/_astro/decision-models-alert-fatigue-perimeter-security-hero.LNxAv7xw_2s9jVu.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Perimeter Security ](/nsights/decision-models-alert-fatigue-perimeter-security/) [Could Decision Models Fix Alert Fatigue in Perimeter Security?](/nsights/decision-models-alert-fatigue-perimeter-security/) [ ![A pole-mounted security camera overlooks a trailer yard at night. One light pole lights a patch of cracked asphalt and the nearest trailers, and the yard fades into darkness toward the back fence.](/_astro/video-analytics-at-night-hero.BexOWClS_gsrD7.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security AI Video Analytics ](/nsights/video-analytics-at-night/) [Video Analytics at Night Sees a Different Image](/nsights/video-analytics-at-night/) [ ![A pole-mounted security camera at dusk looking along a long chain-link fence line that recedes into the distance beside an empty paved yard.](/_astro/pixels-per-foot-video-analytics-hero.C9jC7W5__ADrsL.webp) _![Eye Icon](/_astro/nsightify-icon.BQ_K2r-1_ZqjMsD.svg)_ AI & Security Camera Integration ](/nsights/pixels-per-foot-video-analytics/) [How to Calculate Pixels per Foot for Video Analytics](/nsights/pixels-per-foot-video-analytics/) See Nsightify in Action We're onboarding a limited number of pilot partners. If you're an operations or security leader in construction, warehousing, or manufacturing — let's talk. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## We're Onboarding Pilot Partners. URL: https://nsightify.com/pilot-program/ Nsightify is onboarding a limited number of pilot partners — early access to AI video analytics, dedicated onboarding, preferred pricing, roadmap input. Pilot Program [Home](/) \> Pilot Program # We're Onboarding Pilot Partners. Nsightify is in controlled early access. We're working with a select group of operations and security leaders to deploy, validate, and refine the platform before full commercial launch. If your business has active cameras and a real problem to solve — we want to hear from you. [Apply to Join the Pilot](/contact/?interest=pilot) What you get ## What Pilot Partners Receive Dedicated Onboarding & Support Every pilot partner works directly with our team through setup, configuration, and first deployment. We’re available throughout — not just at kickoff. Early Access Pricing Pilot partners lock in preferred pricing before public launch. The rate reflects the role you play in helping us build the right product. Direct Roadmap Input Your feedback shapes what we build next. Pilot partners identify real gaps, request specific capabilities, and see their input reflected in the platform’s development. Custom Use Case Support If your operation has a detection requirement that doesn’t fit a standard configuration, we’ll work with you to scope it. Custom builds are part of what early partnership means. Who is the right fit? ## Who We're Looking For * Have existing IP or CCTV cameras in active operation * Work in construction, warehousing, manufacturing, or security * Have a real challenge in physical safety or perimeter security * Can commit a point of contact for ongoing feedback * Are decision-makers — not evaluating on behalf of a procurement committee This is not a free trial — it's a working partnership. The pilot process ## How It Works 01 ### Tell us your challenge Share the specific problem you’re trying to solve. We’ll determine whether Nsightify is the right fit and what a deployment looks like for your environment. 02 ### We scope it together Our team proposes a configuration based on your camera setup, use cases, and deployment preference — Nsightify Cloud or On-Premises (Zero Trust). 03 ### Deploy and validate We go live. You use the platform. We support the full deployment and iterate on what you find. 04 ### Shape what comes next Your feedback feeds directly into Nsightify’s roadmap. Pilot partners see high-priority requests reflected in the platform fastest. Why partner with us ## Why Partner With Nsightify Your feedback drives the roadmap Early partners don’t just use the product — they help determine what it becomes. We ship fast and we listen. No camera replacement We work with the cameras you already have. The only extra piece is a small Edge Relay — and only if your cameras can’t reach the internet on their own. Both deployment options available Nsightify Cloud for convenience. On-Premises (Zero Trust) on your own servers for security-sensitive environments. Pilot partners can run either. No long-term lock-in We earn continued engagement by delivering results — not by contract terms. Have a Challenge Worth Solving? Tell us about your operation. We'll tell you whether Nsightify is the right fit — and what pilot participation looks like. [Apply to Join the Pilot](/contact/?interest=pilot) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Privacy Policy URL: https://nsightify.com/privacy/ How Nsightify collects, processes, secures and protects your personal data across our website and B2B AI video analytics platform. Legal [Home](/) \> Privacy Policy # Privacy Policy How we collect, use, share, and protect your personal information across our website and platform. **Effective** \[INSERT DATE\] **Last updated** \[INSERT DATE\] This Privacy Policy explains what information Nsightify collects, how we use it, who we share it with, and the rights you have over your personal information across our Website and Platform. ## 1\. Introduction Nsightify ("Nsightify," "we," "us," or "our") operates **nsightify.com** (the "Website") and the Nsightify AI video analytics platform (the "Platform"). We are headquartered in Los Angeles, California, United States. This Privacy Policy explains what information we collect, how we use it, who we share it with, and what rights you have over your personal information. It applies to: * Visitors to nsightify.com * Prospective and current business customers and their authorized users who access the Platform * Anyone who contacts us or interacts with our marketing communications If you are a business customer using our Platform, your use is also governed by the agreement between Nsightify and your organization. Please read this policy carefully. By using our Website or Platform, you acknowledge the practices described here. ## 2\. Information We Collect ### 2.1 Information You Provide Directly When you visit our Website or contact us, you may voluntarily provide: * **Contact and inquiry data** — your name, business email address, job title, and company name when you submit a demo request, contact form, or other inquiry * **Newsletter subscription data** — your email address when you sign up for product updates or communications ### 2.2 Information Collected Automatically When you visit our Website, we and our service providers automatically collect certain technical data, including: * **Usage data** — pages visited, time spent on pages, links clicked, and referring URLs * **Device and browser data** — IP address, browser type and version, operating system, and device identifiers * **Cookie and tracking data** — see Section 5 for full details on the cookies and tracking technologies we use ### 2.3 Platform Data — What Nsightify Processes on Behalf of Customers The Nsightify Platform is a business-to-business (B2B) service. When our customers deploy the Platform, it connects to their existing camera infrastructure and processes live video streams to generate real-time operational intelligence. **What Nsightify does NOT collect or store:** * Raw video footage or recordings from customer camera systems * Archives or copies of camera streams * Personally identifiable information about individuals who appear in video feeds **What Nsightify does collect:** * **Anonymized, aggregated analytical data** — our AI models generate statistical output data (such as event counts, alert frequencies, and compliance metrics). This data is non-identifiable and aggregated at the operational level. We may retain this data for model performance monitoring, service improvement, and customer reporting. ### 2.4 Our Role: Data Controller vs. Data Processor For **Website data** (contact forms, newsletter signups, live chat), Nsightify acts as the **data controller** — we determine how and why your information is processed. For **Platform data**, Nsightify acts as a **data processor** — we process video streams on behalf of our business customers, who act as the **data controllers**. This means: * The business customer is responsible for having the appropriate legal basis and privacy notices in place for individuals who appear in their camera footage (employees, visitors, contractors, or members of the public) * Nsightify processes that video data strictly according to the customer's instructions and in accordance with a Data Processing Agreement (DPA) * If you are an individual whose image may have been captured through a customer's deployment of our Platform, please direct any privacy inquiries to that business directly, as they are the data controller responsible for that processing ## 3\. How We Use Your Information We use the information we collect for the following purposes: Purpose Legal Basis Under GDPR Responding to demo requests and inquiries Legitimate interest / pre-contractual steps Sending marketing communications (with consent) Consent Operating and improving our Website Legitimate interest Analyzing Website traffic and user behavior Legitimate interest Providing and improving our Platform services Contract performance Aggregated model performance monitoring Legitimate interest Complying with applicable legal obligations Legal obligation Preventing fraud and maintaining security Legitimate interest / legal obligation We do not use personal information to make automated decisions that produce legal or otherwise significant effects on individuals. ## 4\. How We Share Your Information **Nsightify does not sell your personal information.** We do not share personal information with third parties for their own independent marketing purposes. We share information only in the following circumstances: * **Service providers** — We work with trusted third-party vendors who help us operate our Website and business (see Section 6). These providers are contractually bound to process your data only on our behalf and subject to appropriate security requirements. * **Business customers** — If you are an authorized user accessing the Platform through a business customer's account, that customer may have access to activity and data generated within their account. * **Legal requirements** — We may disclose information if required to do so by law, court order, or government authority, or where we have a good-faith belief that disclosure is necessary to protect the rights, property, or safety of Nsightify, our customers, or others. * **Business transfers** — If Nsightify is involved in a merger, acquisition, restructuring, or sale of assets, your information may be transferred as part of that transaction. We will provide notice before any such transfer takes effect and before your data becomes subject to a materially different privacy policy. ## 5\. Cookies and Tracking Technologies Our Website uses cookies and similar tracking technologies to operate, analyze, and improve the Website, and to support our advertising efforts. ### 5.1 What We Use Cookie Type Purpose Opt-Out Available? **Essential cookies** Required for the Website to function (e.g., session management, security) No — these are strictly necessary **Analytics cookies** Measure traffic and user behavior to help us improve the Website Yes **Marketing / advertising cookies** Used to measure ad performance and deliver relevant advertising Yes ### 5.2 Specific Services We Use **Google Analytics** We use Google Analytics to understand how visitors use our Website. It collects data such as pages visited, session duration, and approximate geographic region. Data is pseudonymized before processing. You can opt out using the [Google Analytics Opt-Out Browser Add-on](https://tools.google.com/dlpage/gaoptout) or by adjusting your cookie preferences. **Meta Pixel** We use the Meta Pixel to measure the effectiveness of advertising on Meta platforms (Facebook and Instagram) and to build audiences for targeted campaigns. If you have a Meta account, your Website activity may be associated with your profile by Meta. You can manage your preferences through [Meta's Ad Preferences](https://www.facebook.com/ads/preferences). **LinkedIn Insight Tag** We use LinkedIn's Insight Tag to track conversions from LinkedIn campaigns, build retargeting audiences, and analyze Website traffic from LinkedIn members. You can opt out through [LinkedIn's privacy settings](https://www.linkedin.com/psettings/guest-controls). ### 5.3 Managing Your Cookie Preferences When you first visit our Website, you will be asked to consent to non-essential cookies. You can update your preferences at any time through our cookie settings \[add link to cookie preference center\]. You may also configure your browser to block or delete cookies at any time. Note that disabling certain cookies may affect the functionality of some parts of the Website. **For users in the EU, EEA, and UK:** We will not activate marketing or analytics cookies without your prior, informed consent. ## 6\. Third-Party Service Providers We currently use the following categories of third-party service providers who may process personal information on our behalf: Provider Service Headquarters Google Website analytics (Google Analytics) United States Meta Advertising pixel and audience targeting United States LinkedIn Advertising pixel and conversion tracking United States \[Email platform — TBD\] Marketing email delivery \[To be completed\] \[Cloud infrastructure — TBD\] Platform hosting and infrastructure \[To be completed\] All service providers are required to implement appropriate data protection measures and to process your data only as directed by Nsightify. ## 7\. Data Retention We retain personal information for as long as necessary to fulfill the purposes described in this policy, or as required by applicable law. Data Type Retention Period Contact form and demo request data \[X\] years from last contact, or until you request deletion Newsletter subscriptions Until you unsubscribe, or \[X\] years of inactivity Website analytics data Up to 26 months (Google Analytics default) Platform aggregated / anonymized data Retained for service improvement; not linked to any individual When personal information is no longer required, we delete it or irreversibly anonymize it using secure methods. ## 8\. International Data Transfers Nsightify is based in the United States. If you access our Website or use our services from outside the United States, your information may be transferred to and processed in the US or other countries where our service providers operate. We recognize that different countries have different levels of data protection. For transfers of personal data from the European Economic Area (EEA), United Kingdom, or Switzerland to countries not recognized as providing an adequate level of protection, we rely on: * **Standard Contractual Clauses (SCCs)** as approved by the European Commission * **UK International Data Transfer Agreements (IDTAs)** for transfers from the United Kingdom * **Adequacy decisions** where applicable By using our Website or Platform, you acknowledge that your information may be transferred to and processed in countries whose data protection laws may differ from those in your jurisdiction. ## 9\. Your Privacy Rights ### 9.1 EEA, UK, and Swiss Residents (GDPR) If you are located in the European Economic Area, United Kingdom, or Switzerland, you have the following rights under applicable data protection law: * **Right of access** — Request a copy of the personal information we hold about you * **Right to rectification** — Request correction of inaccurate or incomplete information * **Right to erasure** — Request deletion of your personal information, subject to certain legal exceptions * **Right to restriction** — Request that we limit processing of your data in certain circumstances * **Right to data portability** — Request a machine-readable copy of the information you have provided to us * **Right to object** — Object to processing based on legitimate interests, including for direct marketing purposes * **Right to withdraw consent** — Where we rely on consent, withdraw it at any time without affecting the lawfulness of prior processing To exercise any of these rights, contact us at **privacy@nsightify.com**. We will respond within **30 days**. We may need to verify your identity before processing your request. If you are not satisfied with our response, you have the right to lodge a complaint with the data protection authority in your country of residence. ### 9.2 California Residents (CCPA / CPRA) If you are a California resident, you have the following rights under the California Consumer Privacy Act (CCPA) and the California Privacy Rights Act (CPRA): * **Right to know** — Request disclosure of the categories and specific pieces of personal information we have collected about you, and how we use and share it * **Right to delete** — Request deletion of your personal information, subject to certain exceptions * **Right to correct** — Request correction of inaccurate personal information we hold about you * **Right to opt out of sale or sharing** — Nsightify does **not** sell personal information and does not share personal information for cross-context behavioral advertising purposes * **Right to limit use of sensitive personal information** — Request that we limit the use of sensitive personal information to purposes permitted under California law * **Right to non-discrimination** — We will not discriminate against you for exercising any of your CCPA/CPRA rights To submit a California privacy request, contact us at **privacy@nsightify.com**. We will respond within **45 days**, with an extension to 90 days when reasonably necessary. ### 9.3 All Other Users Regardless of your location, you may contact us at any time to: * Access, update, or request deletion of information we hold about you * Opt out of marketing communications * Ask questions about how your data is used We will respond to all reasonable requests in good faith and in a timely manner. ## 10\. Marketing Communications If you have opted in to receive marketing communications from us, we may send you product updates, industry content, and company news. You can unsubscribe at any time by: * Clicking the "Unsubscribe" link included in any marketing email, or * Contacting us at **privacy@nsightify.com** Opting out of marketing does not affect transactional or service-related communications, such as responses to your demo requests or support inquiries. ## 11\. Children's Privacy Our Website and Platform are designed for business use and are intended for adults. We do not knowingly collect personal information from individuals under the age of 16. If you believe that a minor has submitted personal information to us, please contact us at **privacy@nsightify.com** and we will delete it promptly. ## 12\. Links to Third-Party Websites Our Website may contain links to third-party websites or services that are not operated by Nsightify. We are not responsible for the privacy practices or content of those sites. We encourage you to review the privacy policy of any website you visit. ## 13\. Security We implement appropriate technical and organizational measures to protect your personal information against unauthorized access, loss, alteration, disclosure, or destruction. These measures include \[encryption of data in transit, access controls, and regular security reviews\]. No method of transmission over the internet or method of electronic storage is completely secure. If you have concerns about the security of your information or if you believe your information has been compromised, contact us immediately at **privacy@nsightify.com**. ## 14\. Changes to This Policy We may update this Privacy Policy from time to time as our business, legal obligations, or services evolve. When we do, we will update the "Last Updated" date at the top of this page. For material changes — those that significantly affect your rights or how we use your data — we will provide additional notice, such as a prominent banner on our Website or a direct email notification to subscribers. Your continued use of our Website or Platform after any update constitutes your acceptance of the revised policy. ## 15\. Contact Us For questions about this Privacy Policy, to exercise your rights, or to submit a data privacy request, contact us at: **Nsightify** \[Company Address — to be completed upon incorporation\] Los Angeles, California, United States **Email:** privacy@nsightify.com _For EU/EEA and UK residents: If Nsightify designates a Data Protection Representative in the EU or UK, their contact details will be published here._ * * * _© Nsightify. All rights reserved._ --- ## Operational Intelligence Across Your Entire Facility URL: https://nsightify.com/solutions/ PPE detection, physical safety and perimeter security on one platform — 28 detections running on the cameras you already own. Cloud or your own servers. Solutions [Home](/) \> Solutions # Operational Intelligence Across Your Entire Facility Twenty-eight detections across physical safety and perimeter security, on one platform, working from your existing camera infrastructure. Every detection in a solution is included and visible from day one — no premium tier, no add-on bundle. No camera replacement. No downtime. Solutions ## Multiple Capabilities. One Platform. Activate the capabilities that matter most to your operation. Deploy individually or all of them. Every solution connects to the same cameras, the same platform, and the same dashboard. Physical Safety ### Physical Safety 12 detections Keep people and vehicles apart, and keep escape routes clear. Built for construction sites, warehouses, and manufacturing floors where manual oversight doesn't scale — and for any site where someone works alone. * PPE detection (hard hats, hi-vis vests, gloves, goggles) * Unauthorized vehicles in pedestrian walkways * Blocked-exit & egress-path detection * Forklift–pedestrian proximity (near-miss) detection * Person-down, running and commotion detection * Continuous compliance logs for audit and reporting [Explore PPE Detection & Physical Safety](/solutions/physical-safety/) Perimeter Security ### Perimeter Security 16 detections Know exactly who enters your site and when. Detect unauthorized entry, tripwire crossings, loitering, fence-climbing, and someone tampering with a camera the moment it happens — not the next morning in a footage review. * Unauthorized entry & boundary-breach detection * Directional tripwire crossing on gates and fence lines * Camera tampering: obstruction, repositioning, frozen feeds * Tailgating at badge-secured doors and turnstiles * Fence-climbing, perimeter pacing & abandoned objects * After-hours coverage on a schedule you set [Explore Perimeter Intrusion Detection](/solutions/perimeter-security/) Roadmap ### More Use Cases Coming Nsightify is a platform, not a point tool. Animal intrusion for solar farms and substations, unattended baggage, fire and smoke, and licence-plate recognition are all in development — and we build custom detections with pilot partners. [Join the Nsightify Pilot Program](/pilot-program/) Roadmap ## What We're Building Next In development Pilot partners shape the order [Become a Pilot Partner](/pilot-program/) ### Animal Intrusion Livestock, wildlife, and strays crossing into a protected area. The clearest near-term win for solar farms, substations, airfields, and agriculture — sites with long perimeters, no staff on them, and a real cost when an animal gets through. ### Unattended Baggage A dedicated bag, backpack, and suitcase detection that replaces today's general abandoned-object detection with something far more precise — and far quieter — for transit hubs, lobbies, and public buildings. ### Fire & Smoke Early visual detection of fire and smoke on cameras that are already watching the space, minutes before a heat or smoke sensor in the ceiling has anything to report. ### Weapon Detection Identification of a visible weapon in a monitored area. Genuinely hard to do well, so we would rather ship it late than ship it unreliable — an alert your team cannot trust is worse than no alert. ### Licence Plates & Vehicle Attributes Plate reads plus vehicle make, colour, and type. This is what turns access monitoring at gates and barriers from a presence alert into a record of exactly which vehicle came and went, and when. ### Queue Length & Retail Analytics Queue and dwell measurement alongside the counting and occupancy detections already shipping — the beginning of a retail and customer-experience solution rather than a safety one. Deployment ## One Platform. Two Deployment Options. Nsightify Cloud ### Centralized Visibility Fully managed and subscription-based. Access dashboards and alerts from anywhere. Ideal for multi-site operations that need centralized visibility. Nsightify On-Premises (Zero Trust) ### Full Data Sovereignty AI runs locally on your own server(s). No video leaves your network. Full data sovereignty. Designed for security-sensitive or compliance-heavy environments. How it fits together ## From Camera to Alert You keep your cameras. The only extra piece is the Nsightify Edge Relay — and only when your cameras can't reach the internet directly. 01 · Capture Your existing cameras Any RTSP / IP / CCTV. No replacement. When needed 02 · Connect Nsightify Edge Relay A pre-configured mini-PC — only if your cameras can't reach the internet. Managed Nsightify Cloud Fully managed. We run the AI for you. or Self-hosted Your own server(s) Zero-Trust, on-premises. Video never leaves your network. 04 · Act Real-time alerts & dashboard Instant notifications and a live operations view. 05 · Record Logs & compliance Event logs, compliance records, events pushed to your systems. Integrations ## Detections Don’t Stay in the Dashboard. Alerts Reach People Where They Already Are A detection is only useful if someone sees it. Alerts leave the platform and land in the channels your team already checks — so nobody has to sit in front of another dashboard waiting for something to happen. Every Event Is Available to Your Own Systems Detections are published as structured events over an API and outbound webhooks. Your developers can pull them into the software that already runs your operation and build exactly the workflow you want — without waiting on us. If the Connection Doesn’t Exist, We Build It No two operations run the same stack, and we don’t pretend a fixed connector list covers yours. Tell us what the events need to reach, and we scope the integration with you as part of onboarding. Not Sure Where to Start? Tell us about your operation and we'll walk you through the right configuration. Most deployments go live within a day. [Talk to the Team](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) --- ## Your Cameras Cover the Perimeter. Now Make Them Act on It. URL: https://nsightify.com/solutions/perimeter-security/ Detect intrusion, tripwire crossings, tailgating, loitering, fence-climbing and camera tampering in real time on the cameras you already own. Perimeter Security [Home](/) \> [Solutions](/solutions/) \> Perimeter Security # Your Cameras Cover the Perimeter. Now Make Them Act on It. Sixteen perimeter detections on the cameras you already own — intrusion, tripwire crossings, loitering, after-hours activity, tailgating, fence-climbing, and someone tampering with a camera. No security team staring at monitors. No incidents discovered in tomorrow's footage review. [Book a Demo](/contact/) [See How It Works](#how-it-works) The problem ## Cameras Don't Prevent Incidents. Intelligence Does. Most operations have cameras covering every gate, fence line, and entry point. But footage is only as useful as someone’s ability to watch it — and no one can watch everything at once. Incidents are typically discovered during post-event review. By then, the damage is done: unauthorized access has occurred, assets have been compromised, or a safety breach has escalated into something worse. Nsightify changes the equation. Instead of recording what happened, it alerts you while it’s happening. Capabilities ## Proactive Detection Across Your Entire Perimeter ![Night perimeter camera view with Nsightify flagging an unauthorized person inside a fenced restricted zone](/_astro/unauthorized-entry-detection.Ddvbt0s-_1iCmlk.webp) ### Unauthorized Entry & Boundary-Breach Detection Nsightify detects unauthorized individuals entering controlled sites, restricted perimeters, or secured areas and generates an immediate alert. Draw virtual boundaries around your property or specific zones — every crossing is flagged in real time with a timestamped event log. ![Directional tripwire drawn across a site gate: Nsightify alerts on the person crossing inbound and ignores the person walking outbound](/_astro/directional-tripwire-crossing.v-qnrmvv_Z1c1FM1.webp) ### Directional Tripwire Crossing Draw a line across a gate, a fence run, a loading bay, or a one-way doorway, and choose which direction matters. Nsightify alerts on someone coming in without alerting on everyone going out. It is the most precise perimeter control available — and the one that generates the fewest alerts nobody needed. ![Nsightify flagging a person who has remained near a perimeter fence line longer than the allowed dwell time](/_astro/loitering-detection.0zmoEe89_Zlozk5.webp) ### Loitering & Suspicious-Lingering Detection Identify when a person or vehicle remains in a zone beyond a set dwell time. Whether it's someone lingering near a fence line, an unmarked vehicle idling by a gate, or activity in a normally empty area — Nsightify alerts before the situation escalates. ![Nsightify raising an after-hours entry alert as a person walks into a scheduled yard zone at 02:14](/_astro/after-hours-zone-entry.C6ZzuD2l_1YHFDN.webp) ### After-Hours Zone Entry After-hours intrusions are among the most costly security incidents. Nsightify watches your yards and entrances on a schedule you set — silent through the working day, alert the moment someone is there at 02:00. That schedule is the point: it stays a night alarm your team trusts instead of a second feed nobody reads. Coverage continues through connectivity outages. ![Nsightify detecting a perimeter camera lens that has been obstructed, shown beside the same camera's clear reference view](/_astro/camera-tampering-detection.gKTauUCs_1xkgLo.webp) ### Camera Tampering Detection A camera that has been blinded protects nothing, and a frozen feed looks exactly like a quiet night. Nsightify flags a lens that has been obstructed, spray-painted, defocused, or repositioned — and a feed that has stopped moving — within seconds. It reads the picture itself rather than the objects in it, so it covers every camera you own regardless of what that camera is pointed at. ![Nsightify flagging one person climbing a perimeter fence and a second whose repeated back-and-forth path along it is detected as pacing](/_astro/fence-climbing-and-perimeter-pacing.BTP39k9V_ZwPpXK.webp) ### Fence-Climbing & Perimeter Pacing Detection The two behaviours that precede a breach. Pacing flags someone moving back and forth along the same stretch of fence — the classic reconnaissance pattern. Fence-climbing flags sustained upward movement against your fence line. Both give your team the chance to respond while the intruder is still outside. Each is calibrated against your own footage during commissioning. ![Nsightify highlighting an unattended bag left in a restricted perimeter zone with a dwell-time alert](/_astro/abandoned-object-detection.B6fuwViq_Z2cT3xw.webp) ### Abandoned-Object Detection Detect when an object — a bag, box, or package — is left unattended in a monitored zone for an extended period. Useful for construction sites, logistics yards, and any area where unattended items may indicate a security threat. ![Nsightify flagging the removal of stored equipment from a monitored perimeter zone](/_astro/asset-removal-detection.DM6VxnYx_Z5RWUG.webp) ### Asset-Removal Detection Know when an object of interest is removed from a defined zone. Whether it's equipment, materials, or inventory — Nsightify alerts you the moment something that should be there isn't, helping prevent theft and unauthorized removal. ![Security officer receiving an Nsightify perimeter breach alert on a rugged tablet at an industrial site](/_astro/perimeter-breach-alert-routing.BaT39m0P_k0BPU.webp) ### Alerts That Reach the Right People, Immediately Perimeter breach alerts route to on-site staff, a security dashboard, or your own systems over an API and outbound webhooks. No alert gets buried in unreviewed footage. The right person knows within seconds. What's included ## Sixteen Detections. All of Them Included. Every detection below ships with Perimeter Security and is visible in your settings from day one. There is no premium bundle and no add-on module. The only difference between the two groups is whether a detection is ready to run now or needs tuning against your footage first. ### Included and ready to run 11 detections All eleven ship with the solution and are configurable from the day your cameras are connected. Some start the moment a zone or line is drawn; the rest are a single toggle away in the camera's settings, because not every detection belongs on every camera. * Intrusion into a restricted zone The anchor detection of the solution. A person or vehicle enters an area you have marked as off-limits — a fenced yard, a plant room, a protected asset — and the alert fires immediately. * Directional tripwire crossing A virtual line across a gate, fence line, loading bay, or one-way doorway, with a direction attached. The most precise perimeter control there is: you are alerted to someone coming in without being alerted to everyone going out. * Loitering Someone stays in a monitored area past the dwell time you set without leaving. Storefronts, ATMs, entrances, and the quiet corners of a yard where nobody has a reason to linger. * After-hours zone entry Zone presence gated by a schedule — anybody in the yard at 02:00. Silent through business hours by design, so it stays the low-noise night alarm rather than a second intrusion feed. * Camera tampering A lens obstructed, spray-painted, defocused, or repositioned, and a feed that has quietly frozen. It watches the picture itself rather than the objects in it, so it protects every camera regardless of what that camera is pointed at. * Tailgating and piggybacking Two or more people passing an access point in quick succession on one credential. The standard control for badge-secured doors and turnstiles, and the gap most access-control systems cannot see on their own. * Restricted-area dwell Loitering's stricter sibling for genuinely restricted space. Where loitering waits, this escalates to critical the moment someone stops moving through and starts standing still. * Crowd and occupancy limits A live headcount for any area, with an alert above the maximum you set. It also feeds a continuous count to your dashboard, so it is an operations metric as much as an alarm. * Stopped vehicle A vehicle standing where nothing should stand — a fire lane, a gate approach, a drop-off lane, a shoulder. Alerts once it has been stationary longer than you allow. * People and vehicle counting Directional counts across any line, reported on an interval. Entry and exit volumes, yard traffic, and gate flow, without a clicker or a spreadsheet. * Person and vehicle classification A rolling record of what each camera is actually seeing. The instrument you use to understand a site's normal baseline before you decide what should raise an alarm. ### Calibrated to your site first 5 detections Also included, and switched on during commissioning once they have been tuned against your own footage. They wait until then on purpose — nobody should be woken at 3am by an alarm nobody has calibrated. * Wrong-way and direction violation Anything travelling against the expected flow: a wrong-way driver, a vehicle reversing through a one-way lane, a person moving against traffic. * Perimeter pacing and reconnaissance Someone walking back and forth along the same stretch of fence. It is the classic pre-incident pattern and almost nothing else on the market looks for it — which is exactly why it deserves tuning against your site before it goes live. * Fence-climbing and wall scaling Sustained upward movement against a reference line drawn along the top of your fence. Early warning while the intruder is still outside, rather than confirmation once they are in. * Abandoned object A bag, package, or piece of equipment left unattended past a threshold. Lobbies, transit areas, and any space where an item nobody claims is a question worth asking. * Removed object A high-value item disappearing from an area you are watching — equipment, stock, artwork. You learn it is gone in seconds rather than at the next stock count. How it works ## Perimeter Intelligence in Three Steps 01 ### Connect Connect Nsightify to your existing perimeter cameras via RTSP. Any IP camera works. No rewiring, no camera replacement. 02 ### Define Draw your perimeter boundaries and restricted zones in Nsightify’s configuration interface. Set alert conditions, time windows, and notification routing for each area. 03 ### Monitor Nsightify watches every camera continuously. The moment a boundary is breached, an alert fires — in real time, not in the morning’s footage review. What setup involves ## Three Things Worth Knowing Before You Start You decide where the boundaries are. Zone detections prompt you with a shape to place on the camera view. Line detections — tripwire crossing, counting, tailgating, fence-climbing — need you to draw the line first, and they stay quiet until you do. It takes a minute per camera and it is the difference between alerts that matter and alerts your team learns to ignore. Any detection can be put on a schedule, not just the after-hours one. Restrict a detection to a shift, a night window, or a weekend, and choose whether evidence is still captured outside it. Commissioning tunes the system to your site. Time windows are set to your timezone, and the early-access detections are calibrated against your own footage before they are switched on — so the first alert you receive is one worth acting on. Deployment ## Choose How You Deploy Nsightify Cloud ### Multi-Site From One Interface Access perimeter alerts and event logs via the cloud dashboard. Manage multi-site perimeter coverage from a single interface. Nsightify On-Premises (Zero Trust) ### Offline-First, On Your Network Run Nsightify entirely on your own server(s). No footage leaves your network. Offline-first operation means coverage continues even during connectivity outages. Ideal for sites with strict data sovereignty requirements. How it fits together ## From Camera to Alert You keep your perimeter cameras. The only extra piece is the Nsightify Edge Relay — and only when those cameras can't reach the internet directly. 01 · Capture Your existing cameras Any RTSP / IP / CCTV. No replacement. When needed 02 · Connect Nsightify Edge Relay A pre-configured mini-PC — only if your cameras can't reach the internet. Managed Nsightify Cloud Fully managed. We run the AI for you. or Self-hosted Your own server(s) Zero-Trust, on-premises. Video never leaves your network. 04 · Act Real-time alerts & dashboard Instant notifications and a live operations view. 05 · Record Logs & compliance Event logs, compliance records, events pushed to your systems. Comparison ## Nsightify vs. the Alternatives Nsightify Legacy VMS Human Monitoring Real-time breach detection Nsightify Yes Legacy VMS No Human Monitoring Inconsistent Automated alerting Nsightify Yes Legacy VMS No Human Monitoring No Works on existing cameras Nsightify Yes Legacy VMS Sometimes Human Monitoring Yes 24/7 coverage Nsightify Yes Legacy VMS Yes Human Monitoring No Timestamped event logging Nsightify Yes Legacy VMS Partial Human Monitoring No Loitering, pacing and climbing detection Nsightify Yes Legacy VMS No Human Monitoring Inconsistent Industries ## Perimeter Security for Asset-Intensive Operations ### Construction Sites Active sites are high-value targets for theft and unauthorized access. Nsightify monitors every perimeter camera in real time — with zero additional hardware investment. ### Warehouses & Distribution Centers Perimeter breaches at logistics facilities create liability, inventory risk, and worker safety issues. Nsightify provides 24/7 coverage across yards, loading docks, and fencing. ### Manufacturing Facilities Protect controlled access areas and facility perimeters from unauthorized entry — with automated real-time alerts rather than manual security rounds. ### Multi-Site Security Operations For security operations managing multiple locations, Nsightify delivers centralized perimeter intelligence across all sites — without requiring teams to actively monitor every feed. FAQ ## Perimeter Security FAQ ### Will Nsightify work with my existing security cameras? Yes. Nsightify connects to your existing RTSP-compatible IP cameras — you never replace them. For Nsightify Cloud, if your cameras or NVR can’t reach the internet, a small Nsightify Edge Relay (software on a regular PC on your network) connects them securely. For On-Premises, Nsightify runs on your own server(s). ### Do I need any hardware besides my cameras? You never replace your cameras. With Nsightify Cloud, the only possible addition is the Edge Relay — and only if your cameras can’t reach the internet directly. With On-Premises (Zero Trust), you provide one or more servers where Nsightify runs locally so video stays on your network. ### What happens if my internet connection drops? With the On-Premises (Zero Trust) deployment, Nsightify is offline-first. Monitoring and alerting continue during connectivity outages, and events sync automatically when the connection returns. ### Can alerts be routed to my existing systems? Yes. Perimeter breach alerts reach on-site staff through the channels your team already uses, so the right person is notified within seconds. Every detection is also published as a structured event over an API and outbound webhooks, so your own systems can consume it directly. Where the connection you need doesn’t exist yet, we scope and build it with you during onboarding. ### Does it run 24/7? Yes. Nsightify monitors continuously, including after-hours, without requiring a person to watch the feeds. ### What can Perimeter Security detect? Sixteen detections ship with Perimeter Security. Eleven are included and ready to run: intrusion into a restricted zone, directional tripwire crossing, loitering, after-hours zone entry, camera tampering, tailgating, restricted-area dwell, crowd and occupancy limits, stopped vehicles, people and vehicle counting, and person and vehicle classification. The remaining five — wrong-way movement, perimeter pacing, fence-climbing, abandoned objects, and asset removal — are calibrated against your own footage during commissioning before they are switched on. ### Can Nsightify tell me if someone tampers with a camera? Yes, and it is included and ready to run from day one. Camera tampering detection flags a lens that has been obstructed, spray-painted, defocused, or repositioned, and a feed that has quietly frozen. It reads the picture itself rather than the objects in it, so it works on every camera regardless of what that camera is pointed at. ### Can it detect tailgating at a badge-secured door? Yes. Tailgating detection flags two or more people passing an access point in quick succession on a single credential — the standard control for badge-secured doors and turnstiles. It ships switched off and takes one toggle per camera, plus a line drawn across the doorway. ### Can I restrict a detection to certain hours? Yes, and not only the after-hours detection. Every detection in Nsightify carries its own active-window schedule, so any of them can be limited to a shift, a night window, or a weekend. You also choose whether evidence is still captured outside that window. See Perimeter Security in Action Book a demo and we'll walk through exactly how Nsightify deploys on your current cameras — and what perimeter coverage looks like for your specific site. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) Monitoring safety compliance as well? See how [PPE detection and hazard-zone monitoring](/solutions/physical-safety/) run on the same cameras and the same dashboard. Further reading: [calibrating dwell time to stop loitering detection false alarms](/nsights/loitering-vs-dwell-time-perimeter-false-alarms/), [whether decision models could fix alert fatigue in perimeter security](/nsights/decision-models-alert-fatigue-perimeter-security/), [why AI video analytics doesn't require new cameras](/nsights/you-dont-need-new-cameras-ai-video-analytics/), and [how AI video analytics turns existing camera feeds into real-time intelligence](/nsights/ai-video-analytics-turn-cameras-into-intelligence/). --- ## Stop Finding Out About Safety Violations After the Fact URL: https://nsightify.com/solutions/physical-safety/ AI PPE detection, forklift near-misses, person-down alerts and blocked-exit monitoring on the cameras you already own. Cloud or fully on-premises. Physical Safety [Home](/) \> [Solutions](/solutions/) \> Physical Safety # Stop Finding Out About Safety Violations After the Fact Twelve safety detections across every camera you own — PPE violations, vehicles in pedestrian walkways, forklift near-misses, a worker down, blocked exits, and commotion, in real time. Choose Nsightify Cloud, or On-Premises (Zero Trust) where not a single frame leaves your network. [Book a Demo](/contact/) [See How It Works](#how-it-works) OSHA issued over **$25M+** in workplace safety penalties last year. Head protection and fall protection remain among the top 10 most cited violations every single year. The problem ## Manual Safety Checks Don't Scale Safety managers can’t be in every corner of a job site, warehouse floor, or production facility at once. PPE violations go unnoticed between walkthroughs. Incidents get documented — but rarely prevented. Adding AI shouldn’t force a trade-off on where your footage lives. Some teams want a fully managed cloud; others can’t let sensitive footage leave the premises at all. Nsightify monitors every camera, every frame, continuously — and you choose the deployment. With On-Premises (Zero Trust), no footage ever leaves your network. Either way, no manual oversight required. Capabilities ## Real-Time PPE Detection. Instant Alerts. Full Audit Trail. ![Warehouse camera feed with Nsightify PPE detection flagging a worker missing a hard hat beside a compliant worker in hard hat and hi-vis vest](/_astro/ppe-detection-hard-hat-hi-vis.DDeSoKoy_Z590po.webp) ### PPE Detection: Hard Hats, Hi-Vis Vests, Gloves & Goggles Nsightify identifies workers missing required personal protective equipment — hard hats, high-visibility (hi-vis) vests, gloves, goggles — the moment a violation occurs. Every instance, on every camera, on every shift. Not only the ones a supervisor happened to walk past. ![One camera view spanning two zones under different PPE rules — a loading dock requiring hard hat and hi-vis, an assembly area requiring eye protection — with Nsightify flagging a worker who crossed into the loading dock without a hard hat](/_astro/per-zone-ppe-rules.Ce6eDb2j_1IjkOa.webp) ### Per-Zone PPE Rules One site rarely has one PPE policy. A loading dock, a machine floor, and a visitor corridor each carry their own requirements, and Nsightify enforces each of them independently on the same platform. Draw the zone, set what it requires, and the rule holds without anyone re-stating it at every toolbox talk. ![Nsightify raising a critical alert as a forklift enters a green-marked pedestrian walkway while two workers in the same walkway are tracked as pedestrians](/_astro/vehicle-in-pedestrian-walkway-detection.qVVi0fIH_1G0Eco.webp) ### Unauthorized Vehicles in Pedestrian Walkways Pedestrian–vehicle separation is the most enforceable safety rule on a floor and the hardest one to police. Mark your walkways, crossings, and pedestrian-only areas once, and Nsightify raises a critical alert the moment a forklift, truck, van, or car enters one — every shift, without a spotter posted at the corner. ![Nsightify alerting as a pedestrian enters the proximity zone of an active forklift in a warehouse aisle](/_astro/forklift-pedestrian-proximity-detection.CJrPM7xm_ZlUHPH.webp) ### Forklift–Pedestrian Proximity Detection Detect when a pedestrian comes dangerously close to an active forklift — turning near-misses from close calls into prevented incidents. Coverage runs across every camera, every shift, without a spotter. ![Nsightify exclusion zone drawn around an active robotic cell, raising a breach warning as a worker steps inside it](/_astro/hazard-exclusion-zone-enforcement.CPAbZO31_2hBTwy.webp) ### Hazard-Zone, Exclusion-Zone & Restricted-Area Enforcement Define exclusion zones around active machinery, construction hazards, or high-risk areas. Nsightify alerts you the moment someone enters a zone they shouldn't — and separately when someone lingers in a space they should only be passing through. Before an incident happens, not after. ![Nsightify detecting a loaded pallet blocking the egress path to a marked emergency exit and generating an obstruction alert](/_astro/blocked-exit-egress-detection.BPgX9mHJ_Z2d7cf.webp) ### Blocked Exit & Egress-Path Detection Catch when an emergency exit or evacuation route is obstructed by equipment, pallets, or vehicles. Get alerted in real time so obstructions are cleared before they become a safety violation — or worse. ![Nsightify raising a critical person-down alert for a lone worker lying motionless in an unattended plant room](/_astro/person-down-detection.C7SAZNnk_Z4yhpo.webp) ### Person Down, Fleeing & Commotion Detection The three signals that something has already gone wrong. A worker collapsed and motionless — the difference between being found in seconds and being found at the next round, which is the whole lone-worker problem. Someone running in a space that is normally calm. And sudden erratic movement or a group scattering, the earliest visible sign of an altercation or a panic. ![Four warehouse camera feeds in Nsightify, one healthy and three flagged as defocused, lens-obstructed and frozen](/_astro/camera-tampering-detection.DmcZMefp_Z1LsVNu.webp) ### Camera Tampering Detection Your safety coverage is only as good as the cameras behind it. Nsightify flags a lens that has been obstructed, spray-painted, defocused, or repositioned — and a feed that has quietly frozen — within seconds. It reads the picture itself rather than the objects in it, so it protects every camera you own. ![Nsightify compliance history dashboard listing timestamped PPE violations by zone, camera and severity beside the live camera feed](/_astro/ppe-compliance-audit-log.BpDjp5Xg_ZUkzjp.webp) ### Audit-Ready PPE Compliance Records Every detected violation is logged with a timestamp, zone identifier, and violation type. Access clean PPE compliance reports without pulling footage manually. Your records are always ready — including for an OSHA inspection that opens with a request for three years of injury records. ![Nsightify push notification of a PPE violation in Loading Dock A delivered to a supervisor phone on the warehouse floor](/_astro/safety-alert-routing.CnHCpTog_2v09H8.webp) ### Alerts That Reach the Right People Violations are surfaced immediately — via dashboard, local notification, or as a structured event your own systems consume. Configurable by zone, violation type, and personnel. The right person knows the moment something goes wrong. What's included ## Twelve Detections. All of Them Included. Every detection below ships with Physical Safety and is visible in your settings from day one. There is no premium bundle and no add-on module. The only difference between the two groups is whether a detection is ready to run now or needs tuning against your footage first. ### Included and ready to run 9 detections All nine ship with the solution and are configurable from the day your cameras are connected. Some start the moment a zone is drawn; the rest are a single toggle away in the camera’s settings, because not every detection belongs on every camera. * PPE compliance Anyone in a monitored area without the hard hat, high-visibility vest, gloves, or goggles that area requires. Every instance is caught, on every camera, on every shift — not only the ones a supervisor happens to walk past. * Unauthorized vehicle in a pedestrian zone A forklift, truck, van, or car entering a walkway, crossing, or pedestrian-only area. Pedestrian–vehicle separation is the most enforceable rule on a floor, and this is what enforces it when nobody is watching. * Blocked exit and obstructed egress path An emergency exit, fire lane, or evacuation route blocked by a pallet, a vehicle, or a person. This is the fire-code story: obstructions get cleared while they are still an inconvenience rather than a citation. * Camera tampering A lens obstructed, spray-painted, defocused, or repositioned, and a feed that has quietly frozen. Your safety coverage is only as good as the cameras behind it, so the platform watches its own eyes. * Forklift–pedestrian proximity A pedestrian coming dangerously close to an active forklift. Near-misses stop being close calls nobody logged and start being incidents you prevented. * Person down Someone collapsed or lying motionless in a monitored area. It matters most where people work alone — night shifts, plant rooms, remote corners of a site, eldercare — and turns a discovery at the next round into an alert in seconds. * Crowd and occupancy limits A live headcount for any area with an alert above the maximum you set. Muster points during a drill, control rooms, confined spaces, and anywhere a capacity limit is written into your procedures. * Restricted-area dwell Someone standing in a space they should be passing through — a machine area, a plant room, the wrong side of a guard rail. * Entry into a restricted machine area A person crossing into an exclusion zone around active machinery or a construction hazard. Set the boundary once and it is enforced on every shift. ### Calibrated to your site first 3 detections Also included, and switched on during commissioning once they have been tuned against your own footage. A busy floor and a quiet one look very different to these detections, which is exactly why they are calibrated rather than assumed. * Running or fleeing Unusually fast movement in a space that is normally calm. Brisk walking is fine; someone sprinting away from something is worth knowing about immediately. * Crowd commotion and dispersal Sudden erratic movement or a group scattering at once — the earliest visible indicator of an altercation, a panic, or an unplanned evacuation. * Elopement and wandering A person crossing a containment boundary outbound: a patient leaving a memory-care unit, a child past a property line. A narrow but high-consequence detection for the sites that need it. How it works ## From Camera to Compliance 01 ### Connect Point Nsightify at your existing IP cameras. Supports any RTSP-compatible device. No camera replacement. 02 ### Configure Define your monitoring zones and the safety conditions to enforce — PPE requirements, hazard zones, exit paths, forklift areas. Set different rules for each zone. 03 ### Monitor Nsightify analyzes every frame and delivers real-time violation alerts and compliance logs — continuously. Run it in our cloud, or fully on-premises with no cloud dependency. What setup involves ## Three Things Worth Knowing Before You Start You decide where the boundaries are. Zone detections prompt you with a shape to place on the camera view — the walkway, the exit route, the machine area. Line detections, such as elopement, need you to draw the line first, and they stay quiet until you do. It takes a minute per camera and it is the difference between alerts that matter and alerts your team learns to ignore. Any detection can be put on a schedule. Restrict a detection to a shift, a night window, or a weekend, and choose whether evidence is still captured outside it. A rule that makes sense during production does not have to fire during cleaning. Commissioning tunes the system to your site. Time windows are set to your timezone, and the early-access detections are calibrated against your own footage before they are switched on — because a busy floor and a quiet one look very different to them. Deployment ## Choose How You Deploy Nsightify Cloud ### Multi-Site From One Interface Access safety alerts and compliance records via the cloud dashboard. Manage PPE policies and hazard zones across every site from a single interface. Nsightify On-Premises (Zero Trust) ### Offline-First, On Your Network Run Nsightify entirely on your own server(s). No footage leaves your network. Offline-first operation means safety coverage continues even during connectivity outages. Ideal for sites with strict data sovereignty requirements. How it fits together ## From Camera to Compliance You keep your cameras. The only extra piece is the Nsightify Edge Relay — and only when your cameras can't reach the internet directly. 01 · Capture Your existing cameras Any RTSP / IP / CCTV. No replacement. When needed 02 · Connect Nsightify Edge Relay A pre-configured mini-PC — only if your cameras can't reach the internet. Managed Nsightify Cloud Fully managed. We run the AI for you. or Self-hosted Your own server(s) Zero-Trust, on-premises. Video never leaves your network. 04 · Act Real-time alerts & dashboard Instant notifications and a live operations view. 05 · Record Logs & compliance Event logs, compliance records, events pushed to your systems. Comparison ## Nsightify vs. How Safety Gets Checked Today Nsightify Footage Review Manual Walkthroughs Violation flagged as it happens Nsightify Yes Footage Review No Manual Walkthroughs Inconsistent Covers every camera on every shift Nsightify Yes Footage Review No Manual Walkthroughs No Supervisor alerted within seconds Nsightify Yes Footage Review No Manual Walkthroughs Sometimes Different PPE rules per zone Nsightify Yes Footage Review No Manual Walkthroughs Inconsistent Audit-ready log without pulling footage Nsightify Yes Footage Review No Manual Walkthroughs No Person-down alert for a lone worker Nsightify Yes Footage Review No Manual Walkthroughs Not until the next round Industries ## Continuous Monitoring for Every Kind of Floor ### Construction Job sites are high-risk, high-movement environments with shifting headcounts and changing hazard conditions. Nsightify monitors every active zone — across an entire site, in real time. ### Manufacturing Production floors operate around heavy machinery, chemical exposure, and strict safety protocols. Nsightify enforces compliance continuously without interrupting production. ### Warehousing & Logistics Forklift traffic, loading docks, and high-density worker zones require constant coverage. Nsightify keeps every corner of your operation monitored — at scale. ### Energy & Utilities High-consequence environments where PPE is non-negotiable, and where lone workers are the norm rather than the exception. Nsightify enforces compliance at every monitored zone and raises a person-down alert in seconds. ### Healthcare & Eldercare Person-down detection for residents and lone staff, occupancy limits for control rooms and confined spaces, and elopement alerts when someone crosses a containment boundary they should not — a memory-care unit, a ward door, a property line. FAQ ## PPE Detection & Physical Safety FAQ ### What is PPE detection software? PPE detection software uses computer vision to read your existing camera feeds and identify people who are missing required personal protective equipment — a hard hat, a high-visibility vest, gloves, or goggles — at the moment the violation occurs. Instead of a supervisor catching it on a walkthrough, the system flags every instance across every camera continuously and logs it with a timestamp and zone. ### Can AI detect hard hats and hi-vis vests on existing CCTV cameras? Yes. Nsightify performs hard hat and hi-vis detection on the RTSP-compatible IP and CCTV cameras you already own — no specialised or on-camera analytics hardware is required. The feed is analysed either in Nsightify Cloud or on your own servers, depending on the deployment you choose. ### Which PPE types can Nsightify detect? Common equipment including hard hats, high-visibility (hi-vis) safety vests, gloves, and goggles, configurable by zone so each area enforces its own requirements. A loading dock, a machine floor, and a visitor corridor can each run a different PPE policy on the same platform. ### Do I need to buy new cameras? No. Nsightify works with the RTSP-compatible IP and CCTV cameras you already have — you never replace them, and there is no rewiring. ### Do I need any hardware besides my cameras? You keep your cameras. With Nsightify Cloud, the only possible addition is a small Edge Relay (software on a regular PC on your network) — and only if your cameras can’t reach the internet directly. With On-Premises (Zero Trust), you provide one or more servers where Nsightify runs locally so video stays on your network. ### Does my footage leave my network? It depends on your deployment. With Nsightify Cloud, video is streamed to the cloud for analysis. With On-Premises (Zero Trust), video never leaves your network — only lightweight metadata such as event descriptions, timestamps, and zone identifiers is transmitted. ### How quickly can we go live? Most deployments are configured and delivering alerts within a day. There is no months-long implementation. ### Beyond PPE, what else can Physical Safety detect? Twelve detections ship with Physical Safety. Nine are included and ready to run: PPE compliance, unauthorized vehicles in a pedestrian walkway, blocked exits and egress paths, camera tampering, forklift–pedestrian proximity, person-down, crowd and occupancy limits, restricted-area dwell, and entry into a restricted machine area. The remaining three — running or fleeing, crowd commotion, and elopement — are calibrated against your own footage during commissioning before they are switched on. ### Can Nsightify detect forklift near-misses? Yes. Forklift–pedestrian proximity detection alerts you the moment a pedestrian comes dangerously close to an active forklift, so near-misses become prevented incidents instead of close calls. ### Can it stop forklifts and trucks entering a pedestrian walkway? Yes, and it is included and ready to run from day one. Mark your walkways, crossings, and pedestrian-only areas once, and Nsightify raises a critical alert the moment a forklift, truck, van, or car enters one. Pedestrian–vehicle separation is the single most enforceable safety rule on a warehouse floor, and this is the detection that enforces it on every shift without a supervisor present. ### Can Nsightify detect a worker who has collapsed? Yes. Person-down detection flags someone lying motionless in a monitored area and alerts immediately — the difference between a lone worker being found in seconds and being found at the next round. It is relevant anywhere people work alone: plant rooms, night shifts, eldercare, and remote areas of a site. ### Can I run different safety rules in different areas? Yes, and most sites do. Every zone carries its own rules, so a loading dock, a machine floor, and a visitor corridor can each enforce different PPE requirements and different detections on the same platform. Every detection also carries its own schedule, so rules can differ by shift as well as by area. ### Can Nsightify send safety events into the systems we already use? Yes. Every detection is published as a structured event over an API and outbound webhooks, and alerts route to the channels your team already checks — so a violation reaches people without anyone watching a second dashboard. If a safety event needs to reach an internal tool we don’t already connect to, we scope that integration with you during onboarding rather than asking you to work around it. Ready to Automate Safety Compliance? Tell us about your site. We'll show you exactly how Nsightify maps to your existing cameras and what real-time compliance monitoring looks like for your operation. [Book a Demo](/contact/) ![Nsightify team collaborating in the operations center](/_astro/nsightify-office-call-to-action.GKW-YIkJ_Z1KWm9L.webp) Also monitoring your site boundary? See how [AI perimeter security detects intrusion, loitering and fence-climbing](/solutions/perimeter-security/) on the same cameras. Further reading: [what PPE detection can verify on camera, and what it cannot](/nsights/ppe-detection-what-cameras-cannot-verify/), [real-time physical safety monitoring on cameras you already own](/nsights/physical-safety-monitoring-real-time-ai/), [what OSHA's warehouse National Emphasis Program means through 2031](/nsights/osha-warehouse-emphasis-program-2026/), and [why AI video analytics doesn't require new cameras](/nsights/you-dont-need-new-cameras-ai-video-analytics/). --- ## Terms and Conditions URL: https://nsightify.com/terms/ The terms governing your access, usage and service agreements for the Nsightify website and AI video analytics platform. Legal [Home](/) \> Terms & Conditions # Terms and Conditions These Terms govern your access to and use of the Nsightify website and, where applicable, the Nsightify AI video analytics platform and related services. **Effective** June 18, 2026 **Last updated** June 18, 2026 These Terms and Conditions ("Terms") govern your access to and use of the Nsightify website located at nsightify.com (the "Website") and, where applicable, the Nsightify AI video analytics platform and related services (collectively, the "Services"). By accessing or using the Services, you agree to be bound by these Terms. If you are accessing the Services on behalf of a company or other legal entity, you represent that you have the authority to bind that entity to these Terms. If you do not have such authority, or if you do not agree with these Terms, you must not access or use the Services. ## 1\. Definitions The following definitions apply throughout these Terms: **"Nsightify," "we," "us," or "our"** means Nsightify \[Inc./LLC — legal entity to be confirmed\], the provider of the Services described herein. **"You" or "Customer"** means the individual or organization accessing or using the Services. **"Platform"** means the Nsightify AI video analytics software, including any cloud-hosted (SaaS) or on-premises (Zero Trust) deployment made available by Nsightify. **"Customer Data"** means any video feeds, footage, images, metadata, or related data submitted to or processed by the Platform by or on behalf of the Customer. **"Aggregated Data"** means de-identified, anonymized data derived from Customer Data that cannot reasonably be used to identify any individual person, site, or organization. **"Pilot Program"** means any limited, trial, beta, or non-commercial access to the Platform made available by Nsightify prior to or in lieu of a formal subscription agreement. **"MSA"** means a Master Services Agreement or equivalent commercial contract separately executed between Nsightify and the Customer governing paid access to the Platform. **"Website"** means nsightify.com and all associated subdomains and web pages operated by Nsightify. ## 2\. Acceptance of Terms By visiting the Website, submitting a demo request or contact form, registering for a Pilot Program, or otherwise accessing the Services, you confirm that: 1. You are at least 18 years of age and have the legal capacity to enter into a binding agreement; 2. You are acting on behalf of a business or in a professional capacity — the Services are not intended for personal, family, or household use; 3. You have read, understood, and agree to be bound by these Terms and our [Privacy Policy](/privacy/). If you do not agree to these Terms, you must immediately discontinue access to and use of the Services. ## 3\. Services Overview Nsightify provides AI-powered video analytics software that enables businesses to extract real-time operational intelligence from existing camera infrastructure — without requiring new hardware or workflow changes. The Services are offered in two deployment models: * **Cloud SaaS:** A fully managed, subscription-based platform hosted by Nsightify. * **Zero Trust / On-Premises:** A private deployment model for security-sensitive or compliance-regulated environments, installed within the Customer's own infrastructure. The specific scope, pricing, support obligations, and service level commitments applicable to paid subscriptions are governed exclusively by a separately executed MSA or Order Form. In the event of any conflict between these Terms and a signed MSA, the MSA shall control with respect to the subject matter it addresses. These Terms govern the Website and establish the baseline rules for any access to the Services not otherwise covered by an MSA. ## 4\. Website Use ### 4.1 Permitted Use You may access and use the Website for lawful purposes consistent with these Terms. Permitted activities include submitting inquiries, requesting product demonstrations, accessing published content, and exploring Nsightify's publicly available resources. ### 4.2 Prohibited Conduct You must not: * Use the Website for any unlawful purpose or in a manner that violates applicable local, state, national, or international law or regulation; * Attempt to gain unauthorized access to any part of the Website or its underlying systems, servers, or networks; * Use automated tools, bots, scrapers, or crawlers to extract Website content without Nsightify's prior written permission; * Submit false, misleading, or fraudulent information through any Website form, inquiry, or communication channel; * Introduce viruses, malware, ransomware, or any other harmful or disruptive code to the Website or its infrastructure; * Reproduce, republish, redistribute, or commercially exploit any Website content without Nsightify's express prior written consent; * Take any action that places an unreasonable load on the Website's infrastructure or interferes with other users' access. ## 5\. Account Registration Certain features of the Services — including access to the Platform — may require you to create an account. By registering for an account, you agree to: * Provide accurate, current, and complete information; * Keep your account credentials confidential and not share them with any third party; * Notify Nsightify promptly at [support@nsightify.com](mailto:support@nsightify.com) if you become aware of any unauthorized access to or use of your account; * Accept full responsibility for all activity that occurs under your account credentials. Nsightify reserves the right to suspend or terminate accounts that violate these Terms, display signs of compromise, or are associated with fraudulent or abusive activity. ## 6\. Pilot Program Terms Nsightify may offer limited Pilot Program access to qualified organizations evaluating the Platform. If you are accessing the Services under a Pilot Program arrangement, the following terms apply in full, in addition to all other applicable provisions of these Terms. ### 6.1 Nature of Pilot Access Pilot Program access is provided strictly **"as-is"** and **"as-available."** Nsightify makes no warranties, representations, or guarantees of any kind — express or implied — with respect to the Platform's performance, uptime, availability, accuracy, or fitness for any particular purpose during the Pilot Program period. ### 6.2 No Service Level Agreement No service level agreement ("SLA") applies during the Pilot Program. Nsightify does not guarantee or commit to any minimum uptime percentage, response time, issue resolution timeline, or performance threshold for Pilot Program participants. Any performance metrics communicated during a Pilot Program are informational only and do not constitute binding commitments. ### 6.3 Right to Modify or Discontinue Nsightify reserves the right, at its sole discretion and at any time, with or without notice, to: * Modify, limit, suspend, or discontinue any feature, functionality, or element of the Platform; * Terminate a Pilot Program in its entirety; or * Revoke a specific participant's access to the Pilot Program. Nsightify shall not be liable to you or any third party for any such modification, suspension, or discontinuation. ### 6.4 No Obligation to Continue Participation in a Pilot Program does not create any obligation on the part of Nsightify to offer continued access to the Platform, a commercial subscription, any future version of the Services, or any specific pricing or terms. ### 6.5 Feedback If you provide Nsightify with feedback, suggestions, bug reports, or ideas regarding the Platform during or after a Pilot Program, you grant Nsightify a perpetual, irrevocable, worldwide, royalty-free license to use such feedback for any purpose without restriction, attribution, or compensation to you. ## 7\. Acceptable Use of the Platform You agree to use the Platform solely for lawful business purposes and in full compliance with all applicable laws and regulations, including but not limited to workplace safety laws, privacy and data protection regulations, employment and anti-discrimination legislation, and export control and sanctions laws. **You must not use the Platform to:** * Monitor, record, or surveil individuals without proper legal authorization or in violation of applicable privacy, wiretapping, or workplace monitoring laws in your jurisdiction; * Collect, process, or store biometric or sensitive personal data without obtaining all legally required consents and authorizations; * Facilitate or enable unlawful discrimination against any individual based on any protected characteristic; * Reverse engineer, decompile, disassemble, or otherwise attempt to derive the source code, underlying algorithms, or architecture of the Platform; * Copy, modify, create derivative works of, or frame any part of the Platform or its underlying software; * Resell, sublicense, distribute, or make the Platform available to any third party without Nsightify's express prior written consent; * Use outputs, data, or learnings from the Platform to develop, train, or improve a product or service that competes with Nsightify; * Circumvent, disable, or otherwise interfere with any security or authentication mechanism of the Platform. ## 8\. Intellectual Property ### 8.1 Nsightify's Intellectual Property All intellectual property rights in and to the Platform, Website, documentation, software, AI models, algorithms, trademarks, trade names, logos, and associated materials are owned exclusively by Nsightify or its licensors. These Terms do not transfer any ownership rights in Nsightify's intellectual property to you. All rights not expressly granted herein are reserved. ### 8.2 License Granted to You Subject to your compliance with these Terms — and, where applicable, the terms of a signed MSA — Nsightify grants you a limited, non-exclusive, non-transferable, non-sublicensable, revocable license to access and use the Platform solely for your internal business operations during the authorized term of your access. ### 8.3 License Granted to Nsightify You grant Nsightify a limited, non-exclusive, worldwide license to access, store, and process Customer Data solely to the extent necessary to provide and improve the Services as described in these Terms and our Privacy Policy. This license does not give Nsightify the right to sell or otherwise commercialize your Customer Data. ## 9\. Data Ownership and Processing ### 9.1 Ownership of Customer Data You retain full ownership of all Customer Data at all times. Nsightify does not claim any proprietary rights in your Customer Data. Nsightify does not sell, rent, or transfer Customer Data to third parties except as necessary to deliver the Services or as required by law. ### 9.2 Aggregated and Anonymized Model Data Nsightify may analyze and derive Aggregated Data from Customer Data for the purpose of improving the Platform's AI models, training machine learning systems, enhancing detection accuracy, and developing new product capabilities. Aggregated Data is fully anonymized and cannot be used to identify any individual, organization, or site. Nsightify retains all intellectual property rights in and to Aggregated Data and any improvements to the Platform derived from it. ### 9.3 Your Responsibility for Data Legality You are solely responsible for ensuring that your collection, submission, and processing of Customer Data through the Platform — including any video footage that may contain images of individuals — complies with all applicable laws and regulations in your jurisdiction, including laws governing video surveillance, biometric data, employee monitoring, and data subject rights. ### 9.4 Data Security Nsightify implements commercially reasonable technical and organizational security measures to protect Customer Data against unauthorized access, disclosure, alteration, and loss. Specific security obligations applicable to paid subscriptions, including any SOC 2, encryption, or incident response commitments, are governed by the applicable MSA. ### 9.5 Data Retention and Deletion Nsightify retains Customer Data only for as long as necessary to provide the Services or as required by applicable law. Upon termination of access, Nsightify will handle Customer Data in accordance with our [Privacy Policy](/privacy/) and any obligations set forth in a signed MSA or Data Processing Agreement. ## 10\. Privacy and Data Protection Your use of the Services is also governed by our [Privacy Policy](/privacy/), which is incorporated into these Terms by reference. Please review our Privacy Policy carefully to understand how Nsightify collects, uses, stores, and protects personal information. ### 10.1 GDPR — European Union and EEA Users If you are located in the European Union or European Economic Area, or if you use the Platform to process personal data of EU/EEA residents, the following applies: * **Controller/Processor Roles:** Nsightify acts as a **data processor** with respect to any personal data contained within Customer Data. You, the Customer, act as the **data controller** and bear full responsibility for establishing a lawful basis for processing under the General Data Protection Regulation (EU) 2016/679 ("GDPR") and any applicable national implementing legislation. * **Data Processing Agreement:** Where required by the GDPR, Nsightify will enter into a Data Processing Agreement ("DPA") with you. To request a DPA, contact us at [privacy@nsightify.com](mailto:privacy@nsightify.com). * **Data Subject Rights:** As the data controller, you are responsible for responding to data subject requests (e.g., access, erasure, rectification) in accordance with GDPR obligations. Nsightify will reasonably assist you in fulfilling such requests to the extent they relate to Customer Data processed by the Platform. * **International Transfers:** Where Customer Data is transferred outside the European Economic Area, such transfers will be conducted in accordance with applicable data transfer mechanisms, including Standard Contractual Clauses where required. ### 10.2 CCPA — California Residents To the extent applicable, Nsightify complies with the California Consumer Privacy Act ("CCPA") as amended by the California Privacy Rights Act ("CPRA"). Nsightify does not sell personal information. Further details on your rights as a California resident are set out in our [Privacy Policy](/privacy/). ### 10.3 Video Surveillance and Monitoring Laws Laws governing video surveillance, recording, biometric data collection, and workplace monitoring vary significantly by jurisdiction and industry. You are solely responsible for ensuring that your deployment and use of the Platform complies with all applicable surveillance, privacy, labor, and data protection laws in every jurisdiction in which you operate. ## 11\. Confidentiality In connection with the Services, each party may receive or have access to the other party's confidential information ("Confidential Information"), including but not limited to business plans, technical specifications, pricing, customer data, and proprietary processes. Each party agrees to: * Hold the other party's Confidential Information in strict confidence using at least the same degree of care it uses to protect its own confidential information, and no less than reasonable care; * Use Confidential Information solely for the purposes of these Terms or an applicable MSA; * Not disclose Confidential Information to any third party without the disclosing party's prior written consent, except to employees, contractors, or advisors who have a need to know and are bound by comparable confidentiality obligations. These obligations do not apply to information that: (a) is or becomes publicly available through no fault of the receiving party; (b) was already known to the receiving party prior to disclosure; (c) is independently developed by the receiving party without reference to the Confidential Information; or (d) is required to be disclosed by law, court order, or regulatory authority, provided that the receiving party gives the disclosing party prompt written notice and cooperates in seeking a protective order where possible. ## 12\. Disclaimers and Limitation of Warranties ### 12.1 As-Is Basis EXCEPT AS EXPRESSLY SET FORTH IN A SEPARATELY EXECUTED MSA, THE SERVICES ARE PROVIDED ON AN "AS-IS" AND "AS-AVAILABLE" BASIS WITHOUT WARRANTY OF ANY KIND, WHETHER EXPRESS, IMPLIED, STATUTORY, OR OTHERWISE. NSIGHTIFY EXPRESSLY DISCLAIMS ALL WARRANTIES, INCLUDING BUT NOT LIMITED TO WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, TITLE, NON-INFRINGEMENT, AND ACCURACY. ### 12.2 No Guarantee of Detection Outcomes Nsightify does not warrant that the Platform will detect, identify, or alert to every incident, safety violation, unauthorized access event, or security threat. The Platform is designed to support and enhance operational decision-making — it does not replace trained personnel, human judgment, or established safety, security, and compliance protocols. Customers remain fully responsible for their own safety and security programs. ### 12.3 Third-Party Systems and Infrastructure The Platform may operate alongside or integrate with third-party camera systems, network infrastructure, cloud providers, or software. Nsightify is not responsible for the performance, compatibility, availability, or security of any third-party systems or services. ## 13\. Limitation of Liability TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW: **13.1** IN NO EVENT SHALL NSIGHTIFY, ITS OFFICERS, DIRECTORS, EMPLOYEES, AGENTS, OR AFFILIATES BE LIABLE FOR ANY INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL, EXEMPLARY, OR PUNITIVE DAMAGES OF ANY KIND, INCLUDING BUT NOT LIMITED TO LOSS OF PROFITS, LOSS OF REVENUE, LOSS OF DATA, LOSS OF BUSINESS OPPORTUNITY, OR BUSINESS INTERRUPTION, ARISING OUT OF OR RELATED TO THESE TERMS OR THE SERVICES — REGARDLESS OF THE LEGAL THEORY ASSERTED AND EVEN IF NSIGHTIFY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. **13.2** NSIGHTIFY'S TOTAL CUMULATIVE LIABILITY TO YOU ARISING OUT OF OR RELATED TO THESE TERMS OR THE SERVICES SHALL NOT EXCEED THE GREATER OF: (A) THE TOTAL FEES ACTUALLY PAID BY YOU TO NSIGHTIFY IN THE TWELVE (12) CALENDAR MONTHS IMMEDIATELY PRECEDING THE EVENT GIVING RISE TO THE CLAIM; OR (B) ONE HUNDRED U.S. DOLLARS (USD $100), IF NO FEES HAVE BEEN PAID — INCLUDING DURING ANY PILOT PROGRAM PERIOD. **13.3** THE LIMITATIONS IN THIS SECTION APPLY REGARDLESS OF THE FORM OF ACTION, WHETHER IN CONTRACT, TORT (INCLUDING NEGLIGENCE), STRICT LIABILITY, OR ANY OTHER LEGAL OR EQUITABLE THEORY, AND SHALL APPLY TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW. Some jurisdictions do not allow the exclusion or limitation of certain types of damages. In such jurisdictions, Nsightify's liability shall be limited to the maximum extent permitted by law. ## 14\. Indemnification You agree to indemnify, defend, and hold harmless Nsightify and its officers, directors, employees, agents, contractors, and successors from and against any and all claims, liabilities, damages, losses, penalties, and expenses (including reasonable attorneys' fees and legal costs) arising out of or related to: * Your use of the Services in violation of these Terms or any applicable law or regulation; * Your unauthorized or unlawful processing, collection, or storage of personal data, biometric data, or video footage through the Platform; * Your breach of any third-party rights, including intellectual property, privacy, or surveillance rights; * Any claim brought by your employees, contractors, data subjects, or regulators in connection with your deployment or use of the Platform; * Any Customer Data you submit, upload, or process through the Services. Nsightify reserves the right to assume exclusive control over the defense of any matter subject to indemnification, at your expense. ## 15\. Third-Party Links and Services The Website may contain links to third-party websites, tools, or services. These links are provided for convenience only. Nsightify does not endorse, control, or assume any responsibility for the content, privacy practices, availability, or security of any third-party website or service. Your access to and use of third-party services is entirely at your own risk and subject to those parties' own terms and conditions. ## 16\. Modifications to These Terms Nsightify reserves the right to update or modify these Terms at any time at its sole discretion. When we make changes, we will: * Update the "Last Updated" date at the top of this page; and * Provide reasonable advance notice of material changes via the Website or by email to the address associated with your account. Your continued access to or use of the Services after the effective date of any modification constitutes your acceptance of the revised Terms. If you do not agree to the modified Terms, you must discontinue use of the Services immediately. ## 17\. Termination ### 17.1 Termination by You You may stop using the Website at any time. To request termination of Platform access, contact Nsightify at [support@nsightify.com](mailto:support@nsightify.com). ### 17.2 Termination by Nsightify Nsightify may suspend or terminate your access to any or all of the Services immediately, with or without cause, and with or without prior notice, in circumstances including but not limited to: * Violation of any provision of these Terms; * Fraudulent, abusive, harmful, or illegal activity in connection with the Services; * Non-payment of applicable fees under an MSA; * Any other conduct that Nsightify determines, in its reasonable judgment, may expose Nsightify or any third party to legal liability or reputational harm. ### 17.3 Effect of Termination Upon termination of your access for any reason: * Your license to access and use the Platform ceases immediately; * You must delete any locally stored or cached Platform components in your possession; * Nsightify will handle Customer Data in accordance with our Privacy Policy and any applicable MSA terms. Provisions of these Terms that by their nature should survive termination shall continue in full force and effect, including without limitation: Sections 8 (Intellectual Property), 9 (Data Ownership), 11 (Confidentiality), 12 (Disclaimers), 13 (Limitation of Liability), 14 (Indemnification), 18 (Governing Law), and 19 (General Provisions). ## 18\. Governing Law and Dispute Resolution These Terms are governed by and construed in accordance with the laws of the State of California, United States of America, without regard to its conflict of laws principles. Any dispute, claim, or controversy arising out of or relating to these Terms, or the breach, termination, or validity thereof, shall be resolved exclusively in the state or federal courts located in Los Angeles County, California. Each party irrevocably consents to the exclusive personal jurisdiction and venue of such courts and waives any objection on grounds of inconvenient forum. Nothing in this section prevents either party from seeking emergency injunctive or other equitable relief from a court of competent jurisdiction where necessary to prevent irreparable harm pending resolution of a dispute. ## 19\. General Provisions ### 19.1 Entire Agreement These Terms, together with our Privacy Policy and any separately executed MSA, DPA, or Order Form, constitute the entire agreement between you and Nsightify with respect to the Services and supersede all prior and contemporaneous discussions, negotiations, representations, and agreements relating to the subject matter hereof. ### 19.2 Severability If any provision of these Terms is held to be invalid, illegal, or unenforceable under applicable law, that provision shall be modified to the minimum extent necessary to make it enforceable, and the remaining provisions shall continue in full force and effect without interruption. ### 19.3 Waiver No failure or delay by Nsightify in exercising any right, power, or remedy under these Terms shall operate as a waiver of that right, power, or remedy. A waiver of any breach or default does not constitute a waiver of any subsequent breach or default. ### 19.4 Assignment You may not assign, transfer, or delegate any of your rights or obligations under these Terms — whether by operation of law or otherwise — without Nsightify's prior written consent. Any purported assignment in violation of this section is null and void. Nsightify may freely assign these Terms in connection with a merger, acquisition, corporate reorganization, or sale of all or substantially all of its assets. ### 19.5 Force Majeure Nsightify shall not be liable for any failure or delay in performance caused by circumstances beyond its reasonable control, including but not limited to acts of God, natural disasters, epidemics or pandemics, power failures, internet outages, cyberattacks, government actions or restrictions, or failures of third-party service providers. ### 19.6 No Agency or Partnership Nothing in these Terms creates or shall be deemed to create any partnership, joint venture, agency, employment, or fiduciary relationship between you and Nsightify. Neither party has the authority to bind the other in any respect. ### 19.7 Notices Any formal legal notices to Nsightify required or permitted under these Terms should be sent in writing to [legal@nsightify.com](mailto:legal@nsightify.com). Nsightify may provide notices to you via the email address associated with your account or via a prominent notice on the Website. ### 19.8 Language These Terms are written in the English language, which shall be the controlling language for all purposes. Any translation of these Terms is provided for convenience only and has no legal effect. ## 20\. Contact Information If you have any questions, concerns, or requests regarding these Terms, please contact: **Nsightify** \[Street Address\] Los Angeles, California, United States **Email:** [legal@nsightify.com](mailto:legal@nsightify.com) **Website:** nsightify.com For privacy-related inquiries, including GDPR and data subject requests, contact: [privacy@nsightify.com](mailto:privacy@nsightify.com) © 2026 Nsightify. All rights reserved.