Active Assailant Threat Intelligence: The 2026 Briefing on Workplace, Retail, and Public-Venue Incident Patterns, Dwell Times, and Detection Failure Modes

Active assailant threat intelligence is the analytical discipline that physical security architects need most in 2026 and have access to least. The FBI publishes annual incident counts. The Bureau of Labor Statistics publishes workplace homicide totals. The Advanced Law Enforcement Rapid Response Training Center publishes response-time research. None of these sources alone tells a security director where the threat surface concentrates, how incidents unfold in time, and where the detection architecture has to compress latency to change outcomes. This briefing assembles the primary-source data into a single reference for leaders responsible for non-K-12 facilities, the workplaces, retail floors, hospitality venues, healthcare campuses, transit hubs, and public-venue settings that account for the majority of national active assailant incident volume.

The pattern across the FBI, BLS, and ALERRT data is consistent across five-year windows and across location categories. Active assailant events are rare in any given facility but high-consequence when they occur. They concentrate in commerce and open-space settings far more than in education. They resolve faster than law enforcement can typically arrive. The structural detection failure that allows them to inflict casualties is not a failure of cameras to see the threat but a failure of the architecture to convert visibility into response within the time available. The math of survivability runs through the alert-to-response interval, and that interval is now the variable that AI-based detection is engineered to compress.

The Incident Pattern: What the FBI Data Says About Where Active Assailants Strike

The FBI's "Active Shooter Incidents in the United States" series, published annually by the Bureau's Office of Partner Engagement, is the canonical primary source on active-shooter event frequency, location distribution, and casualty totals. The Bureau's 2024 report documented 24 designated incidents resulting in 23 deaths and 83 wounded, a 50 percent decrease in incident count from 2023 but still substantially elevated against the 2014-2018 baseline. The 2023 report documented 48 incidents resulting in 105 deaths and 139 wounded. Across the 2019-2023 five-year window, the FBI designated 229 active shooter incidents, an 89 percent increase over the 121 incidents in 2014-2018. The 2020-2024 window held at 223 incidents, a 70 percent increase over 2015-2019. The shape of the trend is unmistakable: the active assailant threat surface in the United States has expanded by roughly two-thirds in a decade, and the volatility year-over-year does not reverse the underlying secular pattern.

Location distribution is where the threat-intelligence picture diverges sharply from public perception. The FBI 2023 report classified 58 percent of incidents as occurring in "open space" settings (parks, parking lots, roadways, plazas), 29 percent in "commerce" locations (retail stores, restaurants, bars, commercial offices), 6 percent in education, 4 percent in healthcare, and the balance distributed across residence, government, and houses of worship. The FBI 2024 report shifted slightly: 50 percent open space, 17 percent commerce, 17 percent education, 12 percent government, 4 percent house of worship, with no healthcare incidents. Across the rolling five-year window, the dominant pattern holds: open-space and commerce locations together account for the substantial majority of the active assailant threat surface. Education, despite carrying the heaviest media weight when incidents occur, accounts for a single-digit to low-teens percentage of total designated active shooter incidents in any given year. For a sector-specific threat and deployment playbook, see Houses of Worship and Faith Communities: The 2026 AI Physical Security Sector Playbook.

The implication for non-K-12 security architecture is direct. Workplace, retail, hospitality, healthcare, transit, and public-venue environments are not edge cases in the active assailant threat picture. They are the central case. A threat-intelligence assessment that treats schools as the primary target class and other facility types as residual risk inverts the actual incident distribution by an order of magnitude.

Workplace Homicide Data: The Other Half of the Active Assailant Picture

The FBI active shooter dataset captures incidents that meet a specific definition: an individual actively engaged in killing or attempting to kill people in a populated area. It excludes most workplace homicides, which are disproportionately driven by interpersonal disputes, robbery-related events, and domestic violence spillover into the workplace, none of which typically meet the "active shooter" threshold. The Bureau of Labor Statistics' Census of Fatal Occupational Injuries (CFOI) is the authoritative primary source for the broader workplace homicide picture.

The CFOI 2023 release documented 458 workplace homicides in the United States, 8.7 percent of all work-related fatalities and roughly 62 percent of the 740 total fatalities from violent acts. Of those 458 homicides, 373 (81 percent) involved shooting by another person and 33 involved stabbing, cutting, or slashing. The data is industry-coded: protective service occupations (police officers, security guards, correctional officers) carried the highest absolute homicide count, followed by transportation and material moving, retail sales, and food service. Healthcare and social assistance, while showing fewer fatal incidents, is consistently the leader in non-fatal workplace assault rates and is the sector with the most aggressive recent regulatory action via the OSHA General Duty Clause. The CFOI series confirms that women, who make up 8.5 percent of total workplace fatalities, account for 18.3 percent of workplace homicides, a disproportionality consistent with domestic-violence spillover patterns documented in the BJS National Crime Victimization Survey workplace violence supplement.

Combining the FBI active shooter data with the BLS CFOI homicide data produces the threat-intelligence baseline: roughly two dozen designated active shooter events plus several hundred workplace homicides per year, the great majority of which involve firearm use. The threat surface is not exclusively mass-casualty events. It is a continuous distribution from interpersonal escalation in the parking lot to designated active shooter incidents in the lobby. A detection architecture designed against only the tail of that distribution misses the bulk of the actual incident volume that operations leaders have to manage.

LIVE CAM-12 · LOBBY APPROACH

The Detection Failure Modes: Where Existing Architectures Break

The threat-intelligence question that matters more than incident count is failure-mode analysis: which structural features of conventional security architectures convert visible threats into casualty events. The patterns are consistent across after-action reporting from FBI investigations, OSHA workplace violence enforcement records, ASIS case studies, and DHS CISA preparedness analyses.

The first failure mode is detection latency at the human-vigilance layer. The vigilance research from the U.S. Air Force in the 1990s, replicated repeatedly since, established that continuous attention to multi-feed video walls degrades sharply within the first 20 minutes and approaches near-chance detection rates after 40 minutes for low-base-rate events. Active assailant incidents are, definitionally, low-base-rate events for any individual facility. A guard watching a 16-camera feed wall has very low probability of catching a weapon entering one frame within the first 60 seconds. The cameras are functional. The human vigilance layer is the failure point.

The second failure mode is alert-chain latency once detection occurs. The conventional chain runs: someone observes the threat, communicates to the security desk, the desk communicates to leadership and 911, decisions propagate to building occupants. Each handoff adds latency. FBI after-action reporting consistently documents 2-to-3 minute intervals between weapon visibility and first protective action in facilities relying on the conventional chain. Detection may happen within seconds. Response begins minutes later.

The third failure mode is geographic uncoverage. The FBI 2023 location data places 58 percent of incidents in open-space environments, particularly parking lots, roadways, and approach paths to commercial facilities. These zones are commonly the worst-monitored part of the typical facility's camera deployment. Interior cameras are dense in the lobby and main public areas. Exterior coverage of the parking lot, loading dock, rear approach, and walk-up routes is often sparse, low-resolution, recorded-only-not-monitored, or nonexistent. The threat geometry has its highest concentration in the spaces where conventional architectures have their weakest detection coverage.

The fourth failure mode is the integration gap between detection and response systems. Modern facilities have access-control systems, mass-notification platforms, public address infrastructure, and security operations consoles, but these systems are frequently not wired together such that an event in one triggers automated action in the others. A weapon detection alert that surfaces only on a single operator's console, with no automated route to mass-notification, access-control lockdown, or on-shift mobile devices, sacrifices most of the latency advantage that detection technology delivers.

These four failure modes are not technology problems in isolation. They are architecture problems. The cameras work. The radios work. The lockdown systems work. The failure is in the orchestration layer that should be converting detection into coordinated response within the time window where outcomes are still tractable.

The Response Timeline: Where Every Second Has a Calculable Value

The threat-intelligence case for automated detection rests on a single analytically tractable question: how does detection-to-response latency change when human vigilance is replaced by continuous automated analysis, and what does that latency difference mean in terms of survivability outcomes?

The answer is grounded in FBI and DHS data on the pace of active-attack events. The FBI's Active Shooter Incidents reports document, in partnership with the ALERRT Center at Texas State University, that slightly more than half of incidents in the 2000-2022 series resolved before law enforcement arrived on scene. The ALERRT median law-enforcement response time, from the moment of notification, is approximately three minutes; this measure does not include the 60-to-180-second pre-notification interval during which the incident is in progress but no 911 call has yet been placed. The DHS CISA Active Shooter Preparedness materials consistently show that outcomes in the initial 60 to 180 seconds are the primary determinant of total harm.

The response timeline below maps the critical intervals from weapon presentation to law-enforcement arrival, showing where existing architectures accumulate latency and where automated detection compresses it. The figures represent median documented intervals from FBI after-action data, ALERRT response-time research, and the BJS Law Enforcement Management and Administrative Statistics series.

Sector-by-Sector Risk Profile: Where the Threat Surface Concentrates

The aggregate threat surface is useful for executive briefings. For security architecture decisions, the sector-level decomposition is what guides camera placement, alert routing, and response choreography. The sectors below are organized by the empirical incident distribution in FBI, BLS, OSHA, and industry-association data sources, not by anecdotal threat perception.

The Insider Threat Subsurface: Disgruntled-Employee and Domestic-Spillover Patterns

FBI case files and OSHA workplace violence enforcement records consistently show workplace-context incidents have a heavier insider-attacker share than public-venue incidents. The OSHA workplace violence taxonomy classifies incidents as Type I (criminal intent, no relationship to the business), Type II (customer or client), Type III (worker-on-worker), or Type IV (personal relationship, including domestic violence spillover). Type III and Type IV events together account for the majority of workplace homicides where the attacker is identifiable and the motive is documented in the BLS CFOI narrative records and BJS workplace violence supplements.

The implication is that a meaningful share of workplace active assailant events involve attackers with prior facility access, knowledge of building layout, awareness of access control rules, and personal relationships with intended victims. External-threat detection logic, perimeter-focused and stranger-focused, can miss this category entirely. The relevant logic is weapon-presentation logic: the moment a weapon becomes visible in the facility, the source of the threat is no longer relevant to the response decision. The response is the same.

This is the analytical case for treating the weapon detection layer as the operational anchor of the active assailant response architecture. Behavioral intelligence, pre-incident reporting, threat assessment teams, and HR termination-management protocols remain necessary upstream interventions, and a documented workplace violence prevention program is an OSHA regulatory expectation. But the terminal layer, the moment a weapon has crossed the facility line and a casualty event becomes possible within minutes, requires a detection-and-response architecture faster than human vigilance can deliver. That is the architectural niche AI-based weapon detection occupies in a complete program.

The Geographic and Time-of-Day Pattern Within Facilities

Within the workplace and public-venue threat surface, the FBI incident data and the BJS workplace violence series show consistent within-facility geometry. The sub-pattern matters because it determines where camera coverage, AI detection coverage, and physical-security investment have the highest marginal value.

Approach paths and parking are the highest-leverage intervention points. The FBI 2023 location data attributes 58 percent of incidents to open-space environments, with parking lots, roadways, and parking structures forming a substantial subset. The threat geometry typically begins outside the facility envelope. A weapon visible at the parking-lot perimeter at T+0 is the same weapon that is in the lobby at T+30 to T+60 seconds. Detection coverage on approach paths buys the response architecture a window that interior-only coverage cannot replicate.

Lobby and entry-control points are the second leverage zone. The chokepoint geometry is the architectural reason that staffed reception, access-controlled doors, and visitor-management protocols evolved as the dominant entry-security model. AI detection at the lobby threshold, integrated with the access-control system, produces the highest-confidence detection point in most facility floor plans because lighting is controlled, camera angles are designed, and the visitor-density background simplifies the model's discrimination task.

Time-of-day patterns vary by sector. Retail and hospitality concentrate incidents in evening and weekend high-occupancy windows. Healthcare emergency departments concentrate in nights and weekends, when behavioral health volume rises and staffing is leanest. Manufacturing and warehouse facilities carry elevated risk during shift changes and termination meetings, both of which are predictable in time. Sector-specific time-of-day pattern data should drive shift-aligned monitoring posture decisions, not only detection architecture but the staffing of the response side.

What an Adequate Detection Architecture Looks Like in 2026

The threat-intelligence picture above produces a small set of architecture requirements that an active assailant detection program has to meet to credibly compress the response timeline. These are not vendor-specific; they are the structural requirements that any detection program must satisfy regardless of supplier.

First, real-time detection must run continuously across every camera in the high-risk geometry, not on a sampled subset and not only during business hours. Active assailant incidents do not respect monitoring schedules. The FBI 2024 data shows incidents distributed across 19 states and across the 24-hour clock. Coverage gaps in space (uncovered approach paths, rear entries, parking structures) and gaps in time (overnight, weekends, holidays) are the gaps that the threat will exploit.

Second, detection latency must be sub-five-second from frame to alert. The FBI/ALERRT timeline data shows that the difference between a 5-second alert and a 90-second alert is roughly two and a half minutes of compressed lethal window. Five-second detection is the achievable target for production AI weapon detection systems in 2026; anything materially slower than that gives back the latency advantage that justifies the deployment in the first place.

Third, the alert chain integration question is the differentiator between detection systems that change outcomes and detection systems that produce expensive notifications. A system that generates an alert but cannot route it through the building's existing mass-notification platform, public-address system, panic-alert infrastructure, access-control lockdown logic, and on-shift mobile devices adds detection latency back in at the routing step. The integration architecture is not a peripheral consideration. It is the operational core.

Fourth, privacy-by-design architecture eliminates a category of institutional risk that compliance teams, employment counsel, and union representatives rightly flag. Detection systems that operate on object geometry without facial recognition, without biometric matching, and without retained video for detection purposes are deployable in environments where identity-tracking systems are not. For a full technical discussion of how IntelliSee's computer vision platform handles real-time detection without data retention, the technical reference and the Computer Vision Under Occlusion, Low Light, and Adversarial Conditions briefing address the operational questions in depth.

Fifth, response choreography must be rebuilt against the automated detection event. Mass-notification scripts, lockdown logic, evacuation routing, and law-enforcement notification should be rehearsed against a sub-five-second alert, not a human-witness alert. The shift compresses downstream intervals only if the choreography has been rebuilt around it.

How This Threat Intelligence Connects to Operational Decisions

The threat-intelligence picture in this briefing is not a marketing case. It is a planning input. Security architects and risk leaders responsible for non-K-12 facilities should expect to use the FBI, BLS, ALERRT, and DHS data to make several specific decisions in the next planning cycle.

The first is camera and detection coverage prioritization. The 58 percent open-space and 29 percent commerce concentration in the FBI 2023 incident data means that exterior approach paths and commercial-floor entry zones are the highest-leverage coverage areas in most facility footprints. A coverage map that disproportionately weights interior monitoring inverts the empirical threat geometry.

The second is alert-chain audit. The FBI after-action data on response intervals shows that the 90-to-180-second gap between detection and protective action is the structural target for compression. A facility that has invested in detection technology without auditing the alert routing path from detection to lockdown initiation has acquired a data feed, not a response capability.

The third is convergent program design. The OSHA enforcement record, BLS CFOI sector data, and BJS workplace violence supplements all point to the same conclusion: workplace violence prevention is not a single-layer problem. Behavioral threat assessment, visitor management, access control, the weapon detection layer, and response choreography are parts of one program. Funding and reporting structures that treat them as separate programs underinvest in the integration layer where the lethal-window compression happens. The OSHA General Duty Clause Enforcement Reality, the Joint Commission 2026 Workplace Violence Standards, and the DHS SAFETY Act reference are the regulatory complements to this briefing.

The fourth is economic justification. The four-variable ROI model treats threat exposure (incident probability, casualty severity), response compression (minutes saved, alert chain integration), loss-cost economics (workers' compensation, business interruption, reputational impact), and operational footprint (deployment time, false-alert load) as the variables driving the deployment decision. The Four-Variable ROI Framework and the Workers' Compensation Economics brief are the economic complements to this threat-intelligence reference.