A Texas startup is betting that faster, clearer information can change the math of a mass shooting. Angel Protection says its AI software can spot a brandished firearm on ordinary security cameras and push a verified alert to police in under 10 seconds, aiming to close the deadly gap between the first shot and the first 911 call.
Founded in the wake of the Uvalde school shooting, the company’s pitch is simple: use computer vision to detect visible guns before shots are fired, rely on trained human analysts to confirm, and deliver a photo and precise location to first responders so they know exactly whom to look for and where to go.
How Angel Protection’s AI-verified alert system works
Angel Protection integrates with existing CCTV networks in schools, hospitals, and government facilities. Edge devices on-site analyze video feeds multiple times per second to look for firearms. When the model flags a potential gun, a thumbnail and coordinates are routed to a monitoring center in Midland, where human reviewers validate the detection within seconds.
If confirmed, the system sends an alert—with an image of the individual, camera location, and a short clip—to designated security teams and local dispatch. The company emphasizes it does not alert on holstered weapons, focusing on brandishing behavior that indicates imminent harm. It also avoids identity recognition, concentrating solely on object and pose detection.
Hardware quality still matters. Wide fields of view, poor lighting, and low-resolution lenses can reduce performance, particularly for small handguns partially obscured by clothing. Long guns carried openly in parking lots and entryways are generally easier to identify, which is where the company believes it can make the biggest difference.
Racing the 911 clock to shorten police response time
Angel Protection’s internal review of two decades of incidents found that, on average, about 90 seconds pass from the first shot to the first 911 call. Denial and confusion often delay reporting, and when calls do come in, descriptions can be contradictory, leaving officers unsure whom to stop.
The company argues that verified visual intelligence—an image, a location, and a timestamp—cuts through that noise. The goal is to give first responders a head start before a situation cascades. The FBI’s most recent Active Shooter report underscores why speed matters: many incidents unfold quickly, with casualties often occurring before police arrive on scene.
Scope also drives urgency. The Gun Violence Archive has tracked well over 600 mass shooting events annually in the United States in recent years, though definitions vary across organizations. Intervening even half a minute earlier, public safety experts say, can change outcomes in fast-moving attacks.
Privacy safeguards and the risk of false alarms
AI and cameras naturally raise surveillance concerns. Civil liberties groups such as the ACLU and EFF have warned about mission creep and bias in computer vision. Angel Protection says it designed its system to minimize data exposure: processing happens locally, event clips are short, and only human-verified alerts leave the site. There is no face recognition and no continuous streaming to a centralized cloud, the company says.
False positives remain a challenge—especially in open-carry states where rifles may appear in public for lawful reasons. The company has tuned its models to ignore holstered firearms and relies on human reviewers to filter context, but edge cases persist, from ROTC drills to film props and ceremonial events. Accountability logs and after-action reviews are used to refine the models and operating procedures.
A crowded field and the growing need for independent proof
Angel Protection joins a growing ecosystem of AI gun-detection vendors, including ZeroEyes and Omnilert, that pair computer vision with human verification. ZeroEyes has earned a Department of Homeland Security SAFETY Act designation, a liability protection milestone that signals a degree of due diligence. Still, there is no widely accepted federal benchmark for camera-based firearm detection akin to the rigorous tests NIST runs for biometrics, making apples-to-apples comparisons difficult.
Independent validation will be pivotal. Public safety researchers at DHS Science and Technology, the National Institute of Justice, and academic labs have called for standardized testing that measures not only detection accuracy but also false-alarm rates across lighting conditions, crowd density, camera quality, and weapon types. Without common yardsticks, the strongest claims are hard to verify.
For now, Angel Protection says it is monitoring roughly 2,500 cameras in Texas, scanning each feed multiple times per second. Early deployments focus on entrances and parking lots—high-yield locations where attackers often appear with visible weapons before moving inside. The company reports sub-10-second alert times in live environments and says it is working to integrate with 911 dispatch and access-control systems to automate door locks during verified threats.
What to watch next as AI gun detection is deployed
No single technology can end mass shootings, but layers help: trained staff, access control, clear communication, and faster, more accurate alerts. If Angel Protection and its peers can consistently turn minutes into seconds—while respecting civil liberties—they could become part of that stack.
The next markers to watch are third-party evaluations, transparent performance data, and clear policies on retention, auditability, and human oversight. In a field where every second counts and trust is fragile, proof will matter as much as promise.
