AI weapons detection gives a government agency two very different capabilities, and most buyers mix them up. Camera-based AI watches your existing security feeds and flags a firearm the moment someone draws it. Walk-through screening sits at the entrance and finds weapons on people before they get inside. One spots a gun in someone’s hand. The other finds a gun under someone’s jacket. They solve different problems, and the right deployment usually combines them.
Interest from public agencies climbed after high-profile incidents put a spotlight on building entrances and public spaces, and cities have started wiring the technology into existing camera networks. Modesto, California added AI gun detection to municipal cameras in early 2025, and similar municipal and county projects have followed.
Safe and Sound Security secures public agencies nationwide, including the City of Emeryville and the County of Sonoma. This guide explains how the technology works for a government buyer, what it detects, what it costs, and the compliance and procurement steps that decide whether you can actually deploy it.
Key Takeaways
- There are two core technologies, not one. Camera-based (optical) AI detects visible or brandished firearms on your existing cameras. Walk-through sensor systems screen people for concealed weapons at the door. Cameras cannot detect a concealed weapon.
- Compliance is the government gate. NDAA Section 889 bans certain camera manufacturers from government-funded security systems. The camera the AI runs on matters as much as the software, so a non-compliant camera can sink an otherwise compliant project.
- You can buy it through contracts you already use. GSA Schedule, CMAS, Sourcewell, OMNIA Partners, and other cooperative purchasing vehicles let agencies procure from approved contractors without running a full open bid.
- Grants can pay for it. FEMA’s Homeland Security Grant Program awarded $1.008 billion in fiscal year 2025, and the Nonprofit Security Grant Program funds eligible detection equipment.
- Accuracy is uneven, so verification matters. Independent results, including a New York City subway pilot and the Federal Trade Commission’s action against one vendor, show false alarms are real. Human verification and honest scoping beat vendor accuracy claims.
Have a security project?
The Main Types of AI Weapons Detection for Government
Government agencies combine three detection layers, each answering a different question. The first two are what most people mean by “AI weapons detection.” The third complements them.
| Layer | What it detects | Where it fits | Detects concealed? |
|---|---|---|---|
| Camera-based AI (optical) | Visible or drawn firearms on live video | Across an existing camera network | No |
| Walk-through screening (sensor) | Concealed weapons on people | At building entrances | Yes |
| Acoustic gunshot detection | A shot that has already been fired | Wide-area and outdoor sites | No |
Camera-based AI gun detection (optical AI)
Camera-based AI gun detection analyzes your existing camera feeds and alerts staff the moment it sees a drawn firearm. Machine-learning models scan video frames for the shape of a handgun or long gun in someone’s hand, then push a verified alert with the camera location to security staff within seconds. It runs on the cameras you already own, so there is no rip-and-replace.
The hard limit is that camera AI sees only what the camera sees. It detects visible, drawn, or brandished firearms, never a weapon hidden under clothing. Category providers include ZeroEyes, Omnilert, Scylla, IntelliSee, Avigilon, and Spot AI. Safe and Sound Security is brand-agnostic and installs the system that fits the building, the threat profile, and the budget, including Spot AI and other AI camera platforms. For the full mechanics of how optical detection works, see our guide to weapon detection systems.
Walk-through entry screening (sensor-based AI)
Walk-through entry screening finds concealed weapons on people as they enter a building, without making them stop or empty their pockets. Visitors and staff walk through screening lanes at normal pace while millimeter-wave sensors or 3D cameras paired with an AI edge processor scan for concealed metallic and non-metallic threats. This is the technology that replaces the traditional metal detector at a courthouse or city hall lobby.
These systems are built for high foot traffic, which is the point for a busy public building. They are designed to screen for mass-casualty threats such as rifles and larger weapons rather than small close-proximity items. The Department of Homeland Security Science and Technology Directorate surveyed 15 commercially available walk-through systems and noted that they use metal detection or millimeter-wave sensing and are intended for “soft target” sites, while cautioning that its survey did not independently verify vendor performance claims (DHS S&T market survey). Category providers include Evolv, CEIA OpenGate, Xtract One, and SoundThinking. Safe and Sound Security designs and installs these lanes for government security systems and integrates the alerts with the rest of the building’s security.
Acoustic gunshot detection (a complementary layer)
Acoustic gunshot detection identifies a shot after it is fired, which makes it a complement to the other two layers rather than a replacement. Sensors detect the sound signature of gunfire and report a location, which extends coverage to large campuses, parking structures, and outdoor public spaces where entrance screening is impractical. It does nothing to stop a weapon at the door, so agencies pair it with screening or camera AI rather than relying on it alone. See our overview of shooter detection systems for how this layer fits a wider plan.
Visible vs Concealed Weapons: What Each System Actually Detects
Camera-based AI detects only visible, drawn, or brandished firearms. It cannot detect a concealed weapon. A weapon hidden in a bag or under a coat is invisible to a camera until someone pulls it out, at which point the AI can flag it. This is confirmed across the field: camera-based gun detection identifies visible, drawn, or brandished firearms only, not concealed ones.
Concealed weapons require walk-through sensor screening at a controlled entrance. That difference decides which layer an agency needs. If the concern is a person carrying a hidden weapon into a public lobby, the answer is entrance screening. If the concern is a visible threat appearing anywhere across a campus already covered by cameras, the answer is camera AI. Many buyers ask for one when they need the other, so scoping the actual threat first avoids paying for a system that cannot do the job.
Human Verification and the False Alarm Problem
AI flags a possible threat, and a trained person confirms it before anyone responds. That human-in-the-loop step is what keeps the system from drowning staff in false alarms. Leading camera-AI providers route detections to a 24/7 review center, often staffed by military and law-enforcement veterans, or into the agency’s own security operations center. A confirmed alert then sends the location and a dispatch to responders in seconds. ZeroEyes, for example, advertises verified alerts in three to five seconds.
The false-alarm record is the most important thing a government buyer should vet, because the published results are mixed. In a New York City subway screening pilot, sensors logged 2,749 scans that produced 118 false positives and recovered 12 knives but zero guns, and the Legal Aid Society called the program “objectively a failure”. In November 2024 the Federal Trade Commission acted against Evolv Technologies for unsupported claims about its AI screening and allowed certain K-12 customers to cancel multi-year contracts. Schools using these systems have reported repeated false alarms on laptops, binders, water bottles, and instrument cases (Electronic Frontier Foundation analysis).
The root tension is a tradeoff. Raise sensitivity and you catch more weapons but generate more false alarms. Lower it and you reduce nuisance alerts but risk missing a real threat. There is no setting that eliminates both, which is exactly why human verification and professional system design matter. Safe and Sound Security integrates detection alerts into the agency’s existing video management and access control so that a verified alert can trigger a lockdown or mass notification automatically. Security, cabling, and access controls are handled under one contractor, which removes the gaps that a patchwork of separate vendors leaves behind.
Deploying AI Weapons Detection in Government Buildings
Government agencies deploy weapons detection at entrances and across camera networks in courthouses, city halls, DMV and service offices, agency administrative buildings, and transit hubs. The right mix depends on the building. A courthouse or city hall lobby calls for walk-through screening at the public entrance. A large agency campus already covered by cameras is a fit for optical AI across the footprint. Transit stations and open public spaces lean on wide-area camera and acoustic coverage.
Real municipal deployments show the pattern. Modesto, California added AI gun detection to its city camera network in 2025 (GovTech), and renewed federal interest followed high-profile security failures at government facilities (Fast Company). Safe and Sound Security has delivered security systems for public agencies including the City of Emeryville, the City of Laguna Woods, the County of Sonoma, and the California Employment Development Department. For the front-desk side of public-building security, see our guide to visitor management for government.
Privacy and facial recognition
Compliant AI weapons detection identifies the weapon, not the person. Many government installations prohibit facial recognition, and the leading detection systems do not use it or store biometric data. The AI processes video to find a firearm, not to identify or track individuals. This distinction is central to passing a government privacy review.
It is still a real consideration, not a settled one. Civil-liberties groups including the American Civil Liberties Union have raised concerns about government rushing to deploy AI surveillance without oversight. For a public agency, that means a privacy and policy review is part of the buy, not an afterthought. Choosing a system that detects weapons without biometric identification keeps the project on the right side of that review.
Have a security project?
NDAA, TAA, and SAFETY Act Compliance for Government Buyers
For any government-funded deployment, the detection system and the cameras it runs on must clear federal compliance rules. This is where many projects stall, and it is the part vendor marketing tends to skip. Three rules matter most: NDAA Section 889, the Trade Agreements Act, and the DHS SAFETY Act.
NDAA Section 889
Section 889 of the 2019 National Defense Authorization Act bans federal agencies and federal grant recipients from buying or using covered video surveillance equipment from Hytera, Hikvision, and Dahua, including their subsidiaries and OEM rebrands, for the security of government facilities. The ban has applied to grant recipients since August 13, 2020 (CISA; acquisition.gov). Because camera-based AI runs on cameras, a banned camera fails the deployment even when the AI software is fine. Scoping NDAA-compliant cameras is therefore step one. Safe and Sound Security provides NDAA-compliant product options for exactly this reason.
TAA vs NDAA
The Trade Agreements Act and NDAA Section 889 are separate tests, and a product can pass one while failing the other. The TAA governs the country of origin for federal procurement and limits purchases to the United States and designated trade-partner countries. Section 889 bans specific named manufacturers regardless of where a product is made. A camera assembled in a TAA-compliant country can still fail Section 889 if it uses banned components, so government buyers check both.
DHS SAFETY Act tiers
The DHS SAFETY Act gives approved anti-terrorism technologies liability protection, and detection vendors carry different tiers of it. The tier signals how far a product has been vetted, and full Designation can limit a customer’s third-party liability after a covered incident. The levels and example holders are shown below.
| SAFETY Act tier | What it means | Example holders |
|---|---|---|
| Designation | Full approval after testing; broadest liability protection | ZeroEyes, Omnilert |
| Developmental Testing & Evaluation (DT&E) | Provisional status during evaluation | Volt AI |
| Qualified Anti-Terrorism Technology (QATT) listing | Approved for a set term | IntelliSee |
Verify a vendor’s current status directly at the official registry, safetyact.gov, before relying on it, since tiers and dates change. Alongside federal compliance, state and local contracts often require small-business or in-state certifications, and Safe and Sound Security holds California Small Business certification through the Department of General Services.
How Government Agencies Buy and Fund It
Agencies rarely buy detection through a full open bid. Two paths get most deployments done: cooperative purchasing vehicles and federal grants.
Procurement vehicles
Cooperative purchasing vehicles let an agency buy from a pre-approved contractor without running a full RFP. GSA Schedule, CMAS, Sourcewell, OMNIA Partners, BuyBoard, PEPPM, NASPO ValuePoint, DIR, and HGACBuy each hold pre-negotiated pricing and terms that any eligible agency can use. Detection vendors typically sell through these vehicles via integrators and resellers rather than directly. Safe and Sound Security is approved on 15+ government procurement vehicles, which answers the first question most procurement officers ask: can we actually buy this through our system.
Grant funding
Federal grants pay for a large share of public-sector detection deployments. FEMA’s Homeland Security Grant Program awarded $1.008 billion in fiscal year 2025, including about $373.5 million through the State Homeland Security Program (FEMA). The Nonprofit Security Grant Program funds at-risk nonprofit and faith-based organizations (FEMA NSGP). These grants cannot buy weapons or ammunition, but security and detection equipment on FEMA’s Authorized Equipment List is allowable, so scoping a deployment to eligible equipment is what keeps it grant-fundable.
What it costs
Pricing splits into a capital purchase or a subscription, depending on the system. Walk-through screening can be bought outright or leased. A Stadium Tech Report market review put CEIA’s OpenGate at roughly $20,000 for a two-gate system as a capital purchase, while Evolv and Xtract One run on multi-year subscriptions instead. Camera-based AI is usually an annual per-camera software fee on top of the cameras you already own. As of mid-2026, those are the two dominant pricing models, and the right one depends on whether the agency prefers a one-time capital outlay or an operating expense. Safe and Sound Security sizes each deployment to the number of entrances and cameras rather than a fixed package.
Limitations and Honest Considerations
AI weapons detection is one layer of a security program, not a standalone fix. Camera-based AI depends on camera quality, lighting, and placement, and it detects only visible weapons. Walk-through screening covers a controlled entrance but not the rest of the building. Neither replaces trained staff, access control, or an emergency plan.
Vendor claims have outrun results before, which is what drove the FTC action, so a buyer should scrutinize accuracy records rather than marketing. Visible screening can also reassure the public or, handled poorly, signal that a building is unsafe, and privacy review is part of any government buy. Detection works best alongside cameras, access control, mass notification, and staff who know what to do when an alert fires. Safe and Sound Security scopes deployments to what the building actually needs rather than the most expensive option on the shelf.
Also read:
Frequently Asked Questions
Can AI cameras detect concealed weapons?
No. Camera-based AI detects visible, drawn, or brandished firearms only. A weapon hidden in a bag or under clothing is invisible to a camera until it is drawn. Concealed weapons require walk-through sensor screening at a controlled entrance.
Is AI weapons detection NDAA compliant?
It can be, but compliance depends on the specific system and the cameras it runs on. NDAA Section 889 bans covered equipment from Hikvision, Dahua, and Hytera for government-funded security, so a deployment using a banned camera fails even if the AI software is compliant. Government buyers should confirm both the software and the camera hardware.
How much does AI weapons detection cost for a government building?
It depends on the type. Walk-through screening runs into the tens of thousands of dollars per entrance as a capital purchase, such as roughly $20,000 for a two-gate CEIA OpenGate system, or as a multi-year subscription with vendors like Evolv. Camera-based AI is typically an annual per-camera software fee. These figures are current as of mid-2026.
Can federal grants pay for weapons detection?
Yes. FEMA’s Homeland Security Grant Program and Nonprofit Security Grant Program fund eligible detection equipment listed on FEMA’s Authorized Equipment List. The grants cannot buy weapons or ammunition, but security and screening equipment generally qualifies.
Does AI weapons detection use facial recognition?
Leading systems do not. They are built to detect the weapon, not the person, and do not store biometric data. Many government installations prohibit facial recognition, and choosing a system that detects weapons without identifying individuals helps a deployment pass a government privacy review.
The Bottom Line
AI weapons detection works for a government agency when the layer matches the threat. Walk-through screening stops concealed weapons at a courthouse or city hall entrance. Camera-based AI catches a visible firearm anywhere your cameras already reach. Acoustic detection adds outdoor and wide-area coverage. The deployment that actually goes live is the one that also clears NDAA Section 889, carries the SAFETY Act tier you need, gets bought through a procurement vehicle you already hold, and, where possible, gets funded by a homeland security grant.
Safe and Sound Security designs, installs, and maintains AI weapons detection for government agencies nationwide, with NDAA-compliant product options and approval on 15+ government procurement vehicles. Because security, cabling, and access control run under one contractor, a verified weapon alert can trigger lockdown and notification across systems that were built to work together. To scope a deployment for your facility, contact our team.
Have a security project?
Sources
- Department of Homeland Security Science and Technology Directorate: Walk-Through Weapons Screening Systems Market Survey
- Federal Trade Commission: Action Against Evolv Technologies
- Electronic Frontier Foundation: FTC Acts Against Evolv
- CISA: Section 889 Requirements of the 2019 NDAA
- Acquisition.gov: Section 889 Policies
- FEMA: Fiscal Year Homeland Security Grant Program Fact Sheet
- FEMA: Nonprofit Security Grant Program Notice of Funding Opportunity
- DHS SAFETY Act Registry
- GovTech: Modesto Police Add AI to Cameras to Spot Guns
- Fast Company: Capitol Riot Spurs Interest in Gun Detection AI
- American Civil Liberties Union: The Government Is Racing to Deploy AI
- CBS New York: NYC Subway Weapons Detector Pilot Program Ends
- Stadium Tech Report: CEIA OpenGate Scanning System Market Review
