Ninety-five percent of American public schools run active shooter drills. According to the National Academies of Sciences, Engineering, and Medicine, between 10 and 65 percent of students who participate report negative emotional reactions. A 2020 study found a 42 percent spike in anxiety-related language on social media in the days following drills, and a 39 percent jump in depression indicators. That means the single most common school safety exercise in the country is, by the data, causing measurable psychological harm to the children it claims to protect.
Meanwhile, states are starting to ban the most realistic versions of these drills. New York, California, and others have passed laws restricting high-intensity simulations. The National Academies report itself urges every state legislature to prohibit drills that simulate an actual armed assailant in the building. The consensus is forming: the old playbook is broken. But the threat hasn't gone away. So what replaces it?
The answer is shifting from reactive preparation to proactive, real-time AI weapon detection that identifies threats before a shot is fired, without requiring a single child to rehearse their own worst nightmare.
The Drill Paradox: Preparing for Violence by Simulating It
Active shooter drills became standard practice after Columbine in 1999 and accelerated after Sandy Hook in 2012. By the 2015-16 school year, 95 percent of public schools had adopted lockdown procedures. The logic seemed reasonable: if an attack happens, students and staff should know what to do.
But a quarter-century of drilling has produced an uncomfortable truth. As Everytown for Gun Safety found in its research, there is almost no evidence that active shooter drills prevent shootings or meaningfully protect school communities when shootings do occur. What the research does show is harm.
The National Academies' congressionally mandated 2025 report laid out the cost in detail. Nearly one-third of principals and teachers reported being aware of students experiencing trauma or heightened anxiety after drills. Elementary-aged children, who lack the cognitive development to fully distinguish a drill from reality, are especially vulnerable. Students with prior trauma, disabilities, or existing mental health conditions bear a disproportionate burden.
More than 60 percent of youth in one study reported feeling unsafe, scared, helpless, or sad as a direct result of participating in shooter drills. These aren't edge cases. That is the majority of participants telling researchers the experience made them feel less safe, not more.
And the worst part? Only one-third of schools provide any mental health support after conducting drills.
States Are Pulling the Plug on Realistic Drills
Legislators are listening. The trend is clear and accelerating:
New York banned realistic active shooter simulations during the 2024-25 school year, prohibiting exercises that use theatrical props, fake weapons, or actors playing assailants. California's Safe and Prepared Schools Act (AB 1858), effective January 2025, prohibits high-intensity drills that mimic actual shooter scenarios, including theatrical makeup to simulate injuries, individuals posing as attackers, and instructions for students to physically resist assailants.
The National Academies report explicitly recommends that all states follow suit, calling for an end to drills involving deception, high-intensity simulation, or sensory elements designed to mimic a real attack.
But banning harmful drills creates a vacuum. If you stop rehearsing the response, you need something better filling that space. Discussion-based tabletop exercises and standard response protocols are a start, but they are still fundamentally reactive. They assume the attack has already begun. They do nothing to detect the weapon before it's used.
That gap is exactly where AI-powered weapon detection steps in.
Real-Time Detection vs. Rehearsed Response
The core problem with drills is philosophical, not just practical. Drills train people to respond after violence has started. AI weapon detection is designed to identify threats before violence begins. These are fundamentally different security postures.
Camera-based AI weapon detection works by analyzing live video feeds from a school's existing security camera infrastructure. When the system identifies a visible firearm, it generates an alert in real time, typically within seconds. That alert can trigger lockdown protocols, notify law enforcement, and push live visual information to first responders while they're still en route. The critical difference: the system detects the weapon, not the gunshot.
Consider the timeline of a typical school shooting. Research consistently shows these events unfold in minutes. By the time shots are fired, traditional security responses are already behind. A locked classroom door helps, but only if the door was locked before the shooter reached that hallway. Drills practice locking doors after the announcement. AI detection can trigger the announcement before the first shot.
That time gap is where lives are saved or lost.
What the Data Shows About Detection vs. Drills
The statistics on school shootings in America are staggering. Hundreds of incidents have occurred in 2025 and into 2026, and the data shows no sign of the trend reversing. Drills have been nearly universal for over a decade, yet the frequency of school gun violence has not meaningfully declined.
This isn't to say preparation is useless. Staff should know lockdown procedures. Schools should have emergency plans. But the emphasis on drill intensity, the philosophy that making the rehearsal more realistic makes the response more effective, has not been supported by evidence. What has been supported is that detection speed matters enormously.
A study by the Advanced Law Enforcement Rapid Response Training Center at Texas State University found that the average active shooter event lasts roughly 12 minutes, with most casualties occurring in the first few minutes. Every second of advance warning changes the calculus. If a weapon is detected at a school entrance, in a parking lot, or in a hallway before it's discharged, the response window expands from seconds to minutes.
That is the case for investing in detection technology rather than doubling down on drills that harm the people they're supposed to help.
How AI Weapon Detection Actually Works in Schools
There's a practical misconception worth addressing: AI weapon detection doesn't require ripping out existing infrastructure. Systems like IntelliSee work with the security cameras a school already has installed. No metal detectors at every entrance. No airport-style queuing. No additional hardware that creates bottlenecks or makes a school feel like a checkpoint.
Here's the operational flow:
The AI continuously analyzes video feeds across the camera network. It's trained to recognize visible firearms using deep learning models that improve over time. When a weapon is detected, the system generates an alert with the camera location, a visual snapshot, and real-time video. That alert reaches designated security personnel, administrators, and if configured, local law enforcement simultaneously.
This approach also addresses one of the biggest concerns parents and administrators have about school security: privacy. Camera-based AI weapon detection identifies weapons, not people. There's no facial recognition involved. The system doesn't build databases of student identities. It watches for objects, not individuals, which is a critical distinction for school districts navigating privacy regulations and parent concerns.
For school districts evaluating deployment, the practical implementation guide covers camera placement, network requirements, integration with existing emergency notification systems, and staff training protocols.
The Legislation Is Moving in This Direction
While some states ban harmful drills, others are simultaneously mandating AI weapon detection. Georgia's House Bill 1023 requires public schools to implement AI-based weapons detection systems. Multiple states have introduced or passed legislation in 2025 and 2026 creating mandates or funding mechanisms for AI detection technology in schools.
The pattern is consistent: legislators are recognizing that the reactive drill model is insufficient and potentially harmful, and they're looking for proactive technology as the replacement. Security directors who wait for their state to mandate the shift will be playing catch-up. Those who move now get to evaluate technologies on their own timeline rather than scrambling to meet a compliance deadline.
Federal funding is also aligning with this shift. Multiple grant programs now specifically include AI security technology as eligible expenditures, and grant funding resources can offset a significant portion of deployment costs for qualifying districts.
The False Choice Between Drills and Detection
None of this means schools should abandon all emergency preparedness. The National Academies report doesn't recommend eliminating drills entirely. It recommends replacing high-intensity simulations with low-stress, discussion-based exercises and clear, announced procedures.
The most effective school safety posture in 2026 combines three layers: prevention programs that address root causes (threat assessment teams, mental health resources, behavioral intervention), detection technology that identifies weapons before they're used, and simple, humane emergency protocols that prepare staff and students without traumatizing them.
What doesn't belong in that stack is a drill where a teacher barricades a door with a filing cabinet while first graders hide under desks and cry. The data is clear. The legislation is moving. The technology exists.
The security staffing shortage makes this even more urgent. Schools that can't hire enough security personnel can't rely on human surveillance to catch threats. AI fills that gap without requiring additional headcount.
What Security Directors Should Do Now
If you're a security director, superintendent, or facility manager at a K-12 district, here's the actionable path forward:
Audit your current drill practices. Are you running high-intensity simulations? Check whether your state has restricted or banned them. Even if it hasn't yet, the trend suggests it will. Get ahead of it by moving to discussion-based, announced exercises that follow the National Academies' recommendations.
Evaluate your camera infrastructure. Most schools already have security cameras. The question is whether those cameras are just recording footage for after-the-fact review, or whether they can be upgraded with AI analytics that provide real-time threat detection. The difference between passive recording and active detection is the difference between evidence and prevention.
Explore funding. State and federal grant programs increasingly cover AI security technology. Don't assume the budget isn't there before checking what's available.
Prioritize systems that respect privacy. Parents will ask about this. Board members will ask about this. Choose technology that detects weapons without identifying individuals, doesn't use facial recognition, and doesn't create student databases. This isn't a secondary consideration. It's central to community acceptance.
Plan for integration. AI detection is most effective when it connects to your existing emergency notification system, your mass communication platform, and your local law enforcement dispatch. A standalone alert that doesn't reach the right people in the right order doesn't close the response gap.
The Bottom Line
For 25 years, America's primary school safety strategy has been to practice being attacked. The research now shows that practice is damaging the people it's designed to protect, and there is almost no evidence it prevents the attacks it rehearses. States are banning the worst versions of these drills. The National Academies of Sciences, Engineering, and Medicine have called for a fundamental rethinking of the approach.
At the same time, AI weapon detection technology has matured to the point where it can identify a visible firearm on a school campus in real time, using cameras that are already installed, without identifying a single student's face. It shifts the security posture from "what do we do after shots are fired" to "how do we detect the weapon before a shot is fired."
That's not a marginal improvement. That's a different category of protection entirely. And it's one that doesn't require a child to rehearse their own death to work.