Government and Public Buildings: The 2026 AI Physical Security Sector Playbook
How courthouses, municipal offices, and federal facilities are closing the physical security gap with AI-powered computer vision — and what ISC, GAO, and the Improving Federal Building Security Act of 2024 require now.
Government and public buildings face a compounding physical security gap that existing guard-and-camera infrastructure cannot close alone.
The federal building security model has operated on two assumptions for the better part of two decades: that a contracted uniformed guard force plus a camera network constitutes adequate threat detection, and that this combination satisfies both regulatory and risk requirements. The GAO's 2024 covert-test findings — that Federal Protective Service contract guards failed to identify prohibited items in approximately half of 27 test scenarios across federal civilian facilities — put a number on how far those assumptions have drifted from operational reality.
This playbook is for security directors, facility managers, and risk officers responsible for federal agency buildings, courthouses, municipal facilities, state government campuses, and other public-access government environments. It covers where the current detection model breaks down, what primary sources reveal about the threat landscape, how AI-powered computer vision addresses the attention gap without replacing existing infrastructure, and what the procurement and compliance landscape looks like for AI physical security in government settings in 2026.
The federal building security gap: what GAO found and what it means
The Government Accountability Office's 2024 report GAO-25-108085, Federal Building Security: FPS Could Enhance Its Oversight of Contract Guard Forces, is the most consequential primary-source document on government building security published in years. Its findings are specific: across 27 covert penetration tests conducted at federal civilian facilities protected by Federal Protective Service contract guards, approximately 50 percent of tests resulted in testers successfully carrying prohibited items — including simulated explosive components — past security checkpoints without detection.
The GAO identified three structural contributors to this failure rate. First, guard force training and oversight standards were inconsistently applied, particularly for contract personnel rotating across multiple sites. Second, quality assurance inspections by FPS itself were insufficiently frequent relative to the number of facilities and the turnover rates in the contract guard workforce. Third, technical detection technology at many federal buildings remained at a baseline that predates the threat modality evolution of the past decade — X-ray machines and walkthrough magnetometers at entry points, with passive video coverage of the remaining facility.
The third contributor is the one that AI computer vision directly addresses. A guard force scanning a magnetometer queue cannot simultaneously watch the perimeter approach, the parking structure, the building exterior, and the secondary entrances. A camera network records all of these simultaneously but provides no real-time alert unless a human operator is actively watching — and human monitoring attention degrades measurably within 20 minutes on static tasks, as documented by the Mackworth vigilance curve in cognitive psychology research extending back to radar-operator studies from the 1940s and replicated in contemporary video monitoring contexts.
The combination — undertrained guard force plus passive camera coverage — is precisely the gap that AI-assisted detection is designed to close. The GAO's recommendation to FPS to enhance contractor oversight and improve quality assurance is the correct governance response. The technology response is to reduce dependence on sustained human attention as the sole detection mechanism.
The government workplace violence threat landscape: BJS, FBI, and the data behind the risk
Government employees occupy a disproportionate share of workplace violence statistics relative to their share of the overall workforce. The Bureau of Justice Statistics' data on nonfatal workplace violence victimization — drawn from the National Crime Victimization Survey — shows that government and public administration workers experience rates of workplace violence roughly three times higher than private-sector employees across comparable occupational categories. Law enforcement personnel account for a significant portion of this, but corrections officers, court staff, social workers, tax collectors, and public assistance workers all appear at elevated rates.
The FBI's 2024 Active Shooter Report documents 24 confirmed active shooter events across the United States. Three of those — 13 percent — occurred in government, commerce, and public administration settings. That proportion has been relatively stable across the past six years of FBI data, meaning government facilities consistently represent a material share of active shooter events despite representing a fraction of the building stock that commercial facilities do.
Courthouse environments represent a subset of this risk with specific characteristics. The National Council of Juvenile and Family Court Judges has documented courthouse security incidents including the 2019 fatal shooting of a judge at a Steubenville, Ohio courthouse, the 2022 ambush of a Las Vegas judge, and a pattern of weapon interdiction failures at state courthouses that rely on aging magnetometer infrastructure and insufficient staffing for the traffic volumes they process. The Bureau of Justice Statistics has estimated that courthouse and government building employees collectively account for roughly 41 percent of nonfatal workplace violence victims, despite representing approximately 16 percent of the national workforce — a ratio that reflects the public-facing, access-constrained nature of government service environments.
Why passive camera coverage fails when threat density and facility access volumes are both high
A mid-size federal agency building processing 1,500 daily visitors across three entry points generates continuous camera activity across 40 to 80 cameras. A two-person security operations center cannot maintain meaningful active attention on more than 5 to 8 feeds simultaneously, and documented vigilance research shows that accuracy on passive monitoring tasks declines to approximately 60 percent of baseline within the first hour of observation. High-volume, high-stakes environments — courthouse security queues, federal building lobbies, municipal parking structures — require a detection layer that does not fatigue, does not get distracted by legitimate high-traffic activity, and does not require a camera operator to already be watching the right feed when a threat materializes.
Risk profiles across government facility types: federal, state, local, and judicial
Government physical security is not a monolithic problem. The threat surface, access volume, regulatory requirements, and procurement pathways differ significantly across facility categories. A deployment that makes sense for a federal agency building in Washington does not automatically translate to a rural county courthouse or a municipal utility office.
Federal Agency Buildings
The highest regulatory scrutiny and the most developed procurement infrastructure. FPS-managed facilities face ISC Risk Management Process requirements under the Interagency Security Committee standard. AI detection platforms competing for federal installations must navigate FISMA alignment, FedRAMP authorization pathways, and DHS SAFETY Act designation requirements. The threat profile is broad: domestic extremism, workplace grievance events, and opportunistic criminality at public-facing service offices — passport agencies, SSA benefit offices, USCIS interview centers — are all documented. These facilities process high daily visitor volumes with relatively thin staffing ratios.
Federal Courthouses
The highest-acuity government security environment outside of classified facilities. The U.S. Marshals Service operates courthouse security for federal buildings under Article III courts. Threat actors include criminal defendants with active gang affiliations, domestic violence perpetrators, and, historically, individuals targeting judges or prosecutors by name. Weapon interdiction at courthouse entries is the primary security function — but the perimeter approach, the parking structure, and the courthouse exterior all represent pre-magnetometer threat surfaces that passive cameras do not actively monitor.
State Government Campuses
State capitol complexes, agency office parks, and Department of Motor Vehicles networks represent high-access, moderate-threat environments with highly variable security investment. Unlike federal facilities, state buildings often lack a central security oversight body equivalent to FPS or the Marshals Service, resulting in inconsistent detection capability across the same state's footprint. The 2024 Improving Federal Building Security Act applies to federal facilities, but several states have introduced parallel legislation following GAO and Inspector General reports on their own facility security gaps.
State and County Courthouses
The highest-volume judicial security environment in the country. State courts handle 98 percent of all U.S. judicial caseload — roughly 100 million cases annually — across approximately 3,000 courthouse buildings. Staffing ratios are often lower than federal counterparts, magnetometer infrastructure ages faster, and access volumes at busy urban courthouses routinely exceed design capacity during morning docket hours. The NCJFCJ has documented weapon interdiction failures at state courthouses across multiple jurisdictions, typically attributed to staffing gaps at secondary entrances rather than primary screening failures.
Municipal Offices and City Halls
High public-access facilities with significant grievance-motivated threat potential. Planning department hearings, licensing disputes, tax assessment appeals, and benefit determination offices all attract individuals who may be at a high emotional state when they arrive. Municipal facilities rarely have dedicated security forces of the scale deployed at federal counterparts, making AI-assisted detection — which multiplies existing camera coverage without adding headcount — an operationally attractive model. City councils, mayoral offices, and public works departments have all been targets of documented incidents in the FBI's active shooter and workplace violence datasets.
Public Utility and Infrastructure Offices
The intersection of public-access service delivery and critical infrastructure makes utility offices a distinct risk category. Customer service offices for water, power, and transit authorities see significant conflict-driven incidents — disconnection disputes, billing complaints, access denial events — that generate documented assault and threat patterns. The critical infrastructure dimension means that certain utility campuses are also subject to CISA guidance on physical security, adding a compliance layer that purely commercial properties do not face.
AI detection modalities for government environments: what works and where
AI computer vision for government buildings works by analyzing continuous video feeds from existing cameras and applying trained detection models to identify specific threat signatures. The relevant detection modalities for government environments differ somewhat from healthcare settings — the threat surface at a federal building is different from an emergency department — but the underlying architecture is the same: on-premises processing, no cloud video transmission, no facial recognition, no stored video.
AI Detection Modalities Mapped to Government Facility Threat Priorities
| Detection Type | What It Identifies | Primary Government Use Case | Facility Types |
|---|---|---|---|
| Drawn Firearm Detection | Visible, drawn weapon in camera frame — upstream of magnetometer or security queue | Perimeter approach, parking structures, exterior building zones before entry screening | Federal agencies, courthouses, city halls, utility offices |
| Unauthorized Access / Zone Breach | Entry into a defined restricted zone — staff-only corridors, server rooms, judge's chambers approaches, secured parking | Secondary entrance breach detection, tailgating into controlled zones, after-hours intrusion | All government facility types; especially courthouses and federal agencies |
| Loitering Detection | Persistent presence beyond a configured time threshold in a defined zone | Pre-incident surveillance of building approaches, parking deck stairwells, secondary entrances — the pre-attack loitering pattern documented in FBI incident reconstructions | Federal buildings, courthouses, municipal offices, campus perimeters |
| Crowd and Group Formation | Unusual density of individuals gathering in a monitored zone beyond baseline expectations | Protest or demonstration escalation at government facility entrances, lobby confrontations during high-volume service periods | City halls, state capitals, federal service offices, DMV and benefit offices |
| Vehicle Detection and Intrusion | Vehicle presence in pedestrian zones or restricted areas; vehicle approach patterns inconsistent with normal traffic flow | Vehicle-as-weapon threat mitigation at building approaches; parking garage intrusion; after-hours vehicle presence in secured lots | Federal agencies, courthouses with adjacent parking, utility infrastructure campuses |
The detection modality mix for a federal courthouse differs from a county DMV. A courthouse prioritizes drawn-firearm detection at perimeter approaches (upstream of magnetometers), unauthorized access to judicial chambers corridors, and loitering at secondary entrances used by staff. A municipal service office prioritizes crowd formation in waiting areas, loitering at building entrances during off-hours, and zone breach at staff-only service windows. The platform configuration adapts to the facility's specific threat profile; the underlying infrastructure is the same.
One detection capability that is particularly relevant to government settings but rarely discussed explicitly is pre-attack loitering detection. FBI behavioral analysis of active shooter incidents in government facilities — documented in the agency's Making Prevention a Reality series — consistently identifies a pattern of pre-attack location reconnaissance. An individual who visits a facility multiple times in the days before an attack, dwells near exits or security checkpoints, and photographs or maps the layout is exhibiting a pattern that persistent loitering detection can flag for human review. No AI system identifies intent — but it can identify behavioral anomalies that warrant a second look from trained security personnel.
From perimeter detection to coordinated law enforcement dispatch
How AI detection bridges the gap between a camera feed and a dispatched security response at government facilities.
Existing exterior camera captures frame at building approach, parking structure, or secondary entrance — upstream of any magnetometer queue.
Government-sited 1U appliance analyzes frame. No cloud transmission. FISMA posture preserved. Video never leaves the facility network.
CV model identifies drawn firearm, zone breach, or loitering signature. Confidence score evaluated against facility-specific threshold.
Security console, guard radio, facility manager mobile. RapidSOS routes simultaneously to local law enforcement dispatch via 911 integration.
Guard force repositioned. Facility lockdown procedures initiated if threshold met. Law enforcement en route before any human observer confirmed the threat.
Procurement, compliance, and regulatory requirements for AI security in government settings
Government procurement of AI security technology operates through a compliance layer that commercial deployments do not face. Understanding what applies to each facility category is a prerequisite for building a deployment business case that survives internal review.
DHS SAFETY Act designation. The Support Anti-terrorism by Fostering Effective Technologies Act provides liability protection for sellers and buyers of Qualified Anti-Terrorism Technologies. IntelliSee holds DHS SAFETY Act Full Designation — the higher of the two tiers — as a Qualified Anti-Terrorism Technology. For government buyers, SAFETY Act designation is relevant in two ways: it confirms that the technology has passed DHS technical and operational review, and it provides the liability protections outlined in the Act for terrorism-related incidents where the designated technology is deployed. The SAFETY Act designation page covers the specifics of IntelliSee's designation scope.
Interagency Security Committee (ISC) standards. Federal civilian facilities are governed by the ISC Risk Management Process standard, which defines facility security levels (FSL I through V) and mandates baseline countermeasures at each level. AI-assisted video analytics are explicitly recognized as a countermeasure option in the ISC Physical Security Criteria, typically categorized under electronic security systems. A CISA-reviewed AI platform with on-premises processing and no cloud video transmission aligns cleanly with ISC countermeasure requirements for FSL III and IV facilities — the categories that include most mid-to-large federal agency buildings and all federal courthouses.
CISA AI guidance. The Cybersecurity and Infrastructure Security Agency published its AI use in critical infrastructure guidance in 2024, addressing both the opportunity and risk management dimensions of AI deployment in government-adjacent contexts. CISA's framework emphasizes on-premises processing (data sovereignty), minimal data retention, and human-in-the-loop design for high-consequence alert categories — all of which describe IntelliSee's architecture. For state and local government buyers navigating their own compliance landscape, CISA's guidance provides a defensible reference framework even absent a binding federal mandate.
The Improving Federal Building Security Act of 2024. This legislation, introduced in the 118th Congress and advanced with bipartisan support in response to the GAO's findings on FPS performance, directs GSA and FPS to evaluate and implement enhanced technology countermeasures at federal civilian buildings, with specific language around video analytics and automated detection systems. While the Act's implementation regulations continue to develop, it establishes the legislative intent that AI-assisted detection is the direction of federal building security investment — a signal relevant to agencies evaluating whether to accelerate or defer AI security procurement decisions.
Why AI detection without facial recognition is the only viable architecture for most government environments
Government facilities face a specific tension that private commercial buildings do not: the First and Fourth Amendment implications of surveillance technology deployed at public-access government buildings. Courts and legal scholars have raised standing concerns about facial recognition systems at courthouses and government offices that process constitutional rights — from voter registration to immigration hearings. IntelliSee's platform does not perform facial recognition. Detection is based on object-level and behavioral patterns — a drawn firearm, a zone breach, a loitering pattern — not biometric identification. This architectural decision is not simply a product choice; in many government contexts, it is the only architecture that survives legal review. Several states including Illinois, Texas, and Washington have passed biometric privacy statutes that impose strict consent requirements on facial recognition, requirements that are incompatible with government facility security deployment at scale. Object-level AI detection has no such constraint.
What implementation looks like at a government facility
Government facility deployments differ from commercial ones primarily in the integration and approval pathway, not in the underlying technology. A federal agency deployment will navigate IT security review, FedRAMP alignment questions, and potentially a DHS or GSA acquisition vehicle. A county courthouse will navigate local procurement rules. A state utility office may navigate a state IT procurement office. The technology stack underneath these process differences is consistent.
No camera replacement required. The platform connects to existing IP camera networks through the facility's video management system. Most government facilities run established VMS platforms — Milestone XProtect, Genetec Security Center, Avigilon, or comparable systems — all of which support integration. Existing camera investments, wiring infrastructure, and physical security hardware remain in place.
On-premises processing architecture. Detection runs on a dedicated 1U rack-mounted appliance in the facility's own server room. Video does not leave the building network for analysis. For government facilities operating under FISMA, NIST 800-53 controls, or state IT security frameworks, on-premises processing is frequently a prerequisite for approval. It is also operationally resilient — detection continues during external network outages.
Integration with existing dispatch and response systems. Alerts route through the facility's existing guard force dispatch console, to designated security personnel mobile devices, and optionally through RapidSOS directly to local law enforcement CAD systems. Government facilities typically have defined emergency response protocols (equivalent to Code Gray or active-shooter protocols in other sectors) — AI detection sits upstream of these protocols, providing the earliest possible trigger rather than replacing the human response chain.
Deployment timeline. A typical government facility deployment reaches initial detection coverage within 48 to 72 hours of appliance installation. A one- to two-week calibration period follows, during which detection zones are tuned, alert thresholds are calibrated to the facility's specific traffic patterns, and routing is tested against the facility's emergency response workflow. For multi-building government campuses, zone-by-zone rollout over a four- to eight-week period is common.
The economic case for AI physical security at government facilities
Government security investment decisions operate under different constraints than commercial ones — budget cycles are annual, procurement is regulated, and return on investment is often expressed in risk reduction and liability avoidance rather than revenue impact. The economic case for AI-assisted detection in government settings nonetheless rests on measurable variables.
Incident cost avoidance. A single serious active shooter or workplace violence incident at a government facility generates costs that extend well beyond immediate response: federal or state investigation, civil litigation exposure, facility remediation, staff turnover cascades, and long-term reputation effects on the facility's public trust. BLS Census of Fatal Occupational Injuries data shows that fatal workplace violence in government settings carries the same actuarial weight as any other sector — with the added dimension that government employers face both civil liability and political accountability dimensions that private employers do not.
Guard force efficiency multiplier. A government facility deploying AI-assisted monitoring does not reduce its guard force — it redirects it. Guards who previously rotated through passive camera monitoring shifts can be repositioned to active patrols, access control, escort duties, and de-escalation response. This is particularly relevant for FPS contract guard deployments, where the GAO found quality assurance gaps driven partly by the passive, attention-dependent nature of current monitoring tasks.
Insurance and self-insurance actuarial positioning. Many larger government entities self-insure through state or federal risk pools. Risk pool managers are beginning to apply premium differentiation based on documented physical security technology investment — analogous to the commercial insurance market's approach to sprinkler systems and intrusion detection. AI-assisted detection with documented alert logs creates an actuarial record that passive camera systems do not.
Compliance cost avoidance. For facilities subject to ISC or GSA security standards, a documented technology investment that addresses known countermeasure gaps — particularly in the context of GAO or Inspector General findings — reduces the probability of mandatory compliance remediation spending that arrives under less favorable timing and budget conditions.
Government security teams can use IntelliSee's ROI calculator to model these variables against facility-specific parameters, or request a structured risk assessment via the contact page.
Frequently asked questions about AI physical security for government and public buildings
Does AI video analytics technology require replacing existing government camera infrastructure?
No. AI-assisted detection platforms like IntelliSee connect to existing IP cameras through a facility's current video management system. Milestone, Genetec, Avigilon, and most other VMS platforms used in government facilities support integration. A 1U appliance installed in the facility's server room handles all detection processing; no camera replacement, cabling changes, or network re-architecture is required.
How does on-premises AI processing align with government FISMA and cybersecurity requirements?
On-premises processing means video never leaves the facility network. Detection runs on an appliance inside the government building's own server room, which addresses the primary data sovereignty concern under FISMA and NIST 800-53 frameworks. No cloud roundtrip, no external video storage, no third-party data access. For agencies navigating FedRAMP authorization questions, the on-premises architecture separates the detection system from cloud-service authorization requirements.
What is DHS SAFETY Act designation and why does it matter for government procurement?
The DHS SAFETY Act provides liability protection for sellers and buyers of Qualified Anti-Terrorism Technologies. IntelliSee holds SAFETY Act Full Designation — the higher tier — meaning the technology has passed DHS technical review and qualifies for the Act's liability protections. For government procurement offices, SAFETY Act designation is both a technical vetting signal and a liability management factor: deploying a designated technology provides the protections outlined in the Act for terrorism-related incidents.
Does the platform use facial recognition or biometric identification?
No. IntelliSee performs object-level and behavioral detection — drawn firearms, zone breaches, loitering patterns, crowd formation — not biometric identification. This is architecturally significant for government deployments because many states have biometric privacy statutes that impose strict consent requirements on facial recognition, requirements that are structurally incompatible with government facility security applications. Object-level detection carries none of these constraints.
How does AI detection integrate with existing government emergency response protocols?
Alert routing is configurable to existing dispatch workflows. Alerts can go to existing guard force consoles, to security supervisor mobile devices, and optionally through RapidSOS directly to local law enforcement CAD systems. Government facilities with defined active-shooter or workplace violence response protocols integrate AI detection as the earliest-possible trigger — upstream of any human observer — rather than as a replacement for existing response chains.
Can AI detection help address the specific findings in the GAO-25-108085 report on FPS guard force performance?
Yes, in a targeted way. The GAO identified that FPS contract guards failed to detect prohibited items in approximately half of covert tests, and attributed this partly to training gaps and partly to the passive, attention-dependent nature of current monitoring workflows. AI-assisted detection addresses the attention dimension directly: camera feeds are monitored continuously without fatigue or distraction, and specific threat signatures — including weapon presence at perimeter approaches — trigger active alerts rather than relying on an operator to be watching the right feed at the right moment. AI detection does not replace the guard force or address training governance gaps, but it removes the sustained-attention dependency from the detection layer.
What does a government facility AI security deployment typically cost, and are there procurement vehicles available?
Deployment cost scales with camera count, facility count, and detection depth. Government buyers can access AI security technology through GSA Schedule 84 (Law Enforcement and Security Equipment and Services), which simplifies procurement for federal and many state and local agencies. State and local governments may also use cooperative purchasing agreements. Detailed pricing requires a facility assessment, which IntelliSee conducts through its structured risk assessment process — not through a public pricing page, given the security sensitivity of facility-specific configurations.
Continue the research
This playbook covers AI physical security for government and public buildings as a sector. For related research and deployment context:
- Healthcare Workplace Violence: The AI Detection Playbook — companion sector playbook covering AI deployment in hospital environments, including emergency department, behavioral health, and parking structure threat profiles.
- The 2026 Definitive Guide to AI Physical Security — platform-level reference covering how computer vision works across all detection modalities, architecture decisions, and integration pathways.
- Security legislation tracker — actively maintained database of state and federal physical security mandates, including workplace violence prevention statutes and building security legislation.
- AI gun detection — technical reference for the drawn-firearm detection modality, the highest-priority detection type for government facility perimeter monitoring.
- Perimeter control — zone-based unauthorized access detection relevant to courthouse judicial corridors, agency restricted areas, and utility infrastructure campuses.
- DHS SAFETY Act designation — IntelliSee's Full Designation status and what it means for government procurement and liability posture.
More intelligence like this
New IntelliSee research drops monthly at most. Subscribe and get the next sector playbook, technology briefing, or threat intelligence report in your inbox the day it ships.
Request a Risk Assessment
Talk to an IntelliSee security specialist. No sales pitch — a structured conversation about your environment, your threat profile, and whether computer vision is the right fit.
Request a Risk Assessment