AI Is Moving Into the Heart of American Policing, and Almost Nobody Is Watching
From facial-recognition cameras to AI systems that decide where officers go next, a booming industry is selling police departments tools that critics warn could make accountability much harder.
Key points
- The International Association of Chiefs of Police Technology Conference, held in Fort Worth, Texas this past May, showcased AI tools including facial-recognition cameras, drone fleets, gunshot-detection platforms, and automated report-writing software.
- Real-time crime centres, or RTCCs, pull together data from cameras, licence-plate readers, and 911 logs into a single AI-managed dashboard and are now being sold to departments across the country.
- By 2019, the NYPD was collecting roughly two years' worth of body-camera footage every single week, according to a 2019 Committee on Public Safety hearing transcript.
- Axon Enterprise, the company behind the TASER stun gun, acquired surveillance firm Fusus in early 2024 to launch an AI-powered RTCC called Axon Fusus.
- Critics including a fellow at the NYU Center on Race, Inequality and the Law say earlier data-driven policing tools failed to prevent violent encounters, and there's little reason to believe this new wave will be different.
Picture a wall of screens. Each one shows a different camera feed: a street corner, a car park, a school gate. Below them, software quietly decides which patrol car to send where. Nobody voted on this. Most people in the neighbourhood don't know it exists.
That's the product being sold to police departments right now.
The International Association of Chiefs of Police Technology Conference drew thousands to Fort Worth, Texas this past May. The Verge reported that the showroom floor offered facial-recognition cameras, body cameras, chatbots for non-emergency 911 calls, gunshot-detection microphones, drones, and AI report-writers. The pitch was familiar: let the machine handle the paperwork so officers can focus on real work.
Police paperwork isn't really busywork, though. Writing a careful arrest report, reviewing a suspect's history, deciding how to word a charge: these steps sit at the centre of the legal process. Automate them carelessly and real people face real consequences.
The centrepiece of the conference was a category of software called a real-time crime centre, or RTCC. Think of it as an AI control room. It pulls in data from cameras, licence-plate readers, 911 dispatch logs and parole records, then hands officers a summary before they arrive at a scene. The idea is to replace gut instinct with hard data.
New York City's police department pioneered the concept over 20 years ago using human analysts. Data volumes have since grown beyond what any team of humans can track: by 2019, the NYPD was generating roughly two years' worth of body-camera footage every week, according to a 2019 Committee on Public Safety hearing transcript.
Companies including ForceMetrics, with its RTCC product called Velocity, and Axon Enterprise are now competing to supply these systems to departments nationwide. Axon, the company behind the original TASER, bought surveillance firm Fusus in early 2024 and rebranded its RTCC as Axon Fusus. It also sells body cameras, licence-plate readers, an AI report-writer called Draft One, and a drone programme called Axon Air. Critics say the goal is to own the entire technology stack from data collection to decision-making. Motorola Solutions is competing in the same space.
Should ordinary people be worried?
Yes, in a practical sense, because there's almost no federal oversight of how these tools work or what they get wrong. Abrem Ayana, a police captain in Brookhaven, Georgia, told The Verge that much of it amounts to sales gimmicks, yet departments often have to take companies at their word because no independent standards exist. Nina Loshkajian, a fellow at the NYU Center on Race, Inequality and the Law, put it plainly: earlier data-driven policing tools didn't prevent violent encounters between police and the public, and there's no strong evidence the new generation will either.
What's worth watching is whether any of this faces a legal challenge before it becomes too embedded to dislodge. The pattern with past tools, CompStat and PredPol among them, was that they failed visibly and were still kept running for years.
If you live somewhere that uses these systems, ask your local council or police board which AI tools are in use, what data they collect, and who audits them. That question alone tends to get things moving.
