AI police-tech startups race to automate law enforcement decisions

On a visit to Fort Worth, Texas in May 2026, the author attended the International Association of Chiefs of Police Technology Conference to investigate the booming market in AI sales to law enforcement. Press were barred from entering the main venue, but attendees described an exposition floor filled with surveillance and automation products: facial-recognition cameras, license-plate readers, body cameras, AI chatbots to field non-emergency 911 calls, gunshot-detection platforms, drones, and report-writing tools. The underlying pitch, repeated across dozens of vendor booths, echoed the sales language deployed to businesses for years—let machines handle the busywork so humans can focus on meaningful tasks. But in policing, the "busywork" being automated includes critical legal procedures: writing reports, reviewing case histories, and analyzing data to guide resource allocation. These decisions directly affect citizens' lives, and their automation raises profound questions about accountability and bias.

The most sophisticated product category on display was the real-time crime center (RTCC)—a centralized AI system that aggregates data from multiple sources (911 dispatch logs, CCTV feeds, license-plate scanners, body-camera footage) and distills it into actionable insights for patrol officers. Jason Truppi, a former FBI agent specializing in cybercrime who co-founded ForceMetrics in late 2020, explained the appeal: police departments are drowning in data. The NYPD alone was collecting around two years' worth of body-camera footage every week by 2019—far more than any human analyst can meaningfully process. Truppi's company offers Velocity, an RTCC powered by a modified version of ChatGPT, designed to extract patterns and improve "situational awareness" so officers arrive at scenes better informed and less likely to resort to violence. "The creativity is turned down to zero," said Noah Spitzer-Williams, a senior principal product manager at Axon's generative AI division, emphasizing that the system is hallucination-free—a claim the article treats with skepticism, noting that even frontier AI labs like OpenAI, Anthropic, and Google have not solved hallucination entirely. Indeed, Draft One (Axon's report-writing tool) infamously generated a report stating that a Utah officer had "morphed into a frog" after picking up audio from The Princess and the Frog playing in the background.

Axon Enterprise and Motorola Solutions dominate the police-technology stack. Axon, originally known as TASER, is already famous for stun guns and body-worn cameras. In early 2024 it acquired Fusus, a surveillance-technology company, to launch its RTCC offering (Axon Fusus). The company also sells an AI chatbot and Axon Air, a drone program for police. In late 2024, Axon introduced the AI Era Plan, a flat-fee annual subscription granting access to current AI tools and any new ones the company launches. Between the first quarter of 2025 and the first quarter of 2026, AI Era Plan subscriptions surged 140 percent. During an earnings call, Axon President Joshua Isner stated, "We are seeing AI move from early interest to a standard part of how large agencies think about their future technology stack." The company disclosed that AI product revenue grew 700 percent year over year. Police departments typically sign onto these vendors through multiyear contracts and sole-source procurement agreements, which eliminate the need for competing bids when new products are introduced, entrenching the vendor's position.

However, not everyone is sold. Abrem Ayana, a police captain in Brookhaven, Georgia, told the author that "a lot of it is sales gimmicks that don't actually deliver on what the promise is." The concern is deeper than product quality. Nina Loshkajian, a fellow at NYU's Center on Race, Inequality, and the Law, points out that police had already been using predictive algorithms for years before the 2020 push to defund the police, and those systems failed to prevent violent encounters. The track record is instructive: CompStat and PredPol were both early-stage experiments designed to replace fallible human judgment with supposedly objective statistics. Instead, they exacerbated the very problems they were meant to solve. Today's AI systems repeat the same premise—more data, smarter algorithms, less bias—but absent federal oversight or industry standards, police departments have no independent way to verify that vendors' claims are sound. The result is a gold rush, with venture capitalists recognizing police technology as a growth sector and newer startups competing alongside giants like Axon and Motorola to become the de facto AI platform for American law enforcement.

The police technology market is consolidating rapidly around a handful of large vendors who offer bundles spanning data collection, storage, and algorithmic decision-making. Axon and Motorola lock in departments through multiyear contracts, free trials, and sole-source procurement agreements—business practices that reduce competitive pressure and allow vendors to roll out new AI tools without having to re-bid. The influx of venture capital (one-quarter of attendees at the May 2026 conference were equity investors) signals that the sector is viewed as a growth opportunity; Axon's 140 percent year-over-year increase in AI Era Plan subscriptions and 700 percent surge in AI product revenue underscore the appetite. However, the business case rests on a fundamental claim—that AI automation reduces bias and improves decision-making—that prior experience contradicts. CompStat and PredPol were marketed identically as data-driven solutions that would replace fallible human judgment, yet they amplified existing racial inequities. Today's RTCCs operate on the same theory: feed algorithms vast quantities of police data, and they will extract objective patterns. But the data itself often encodes historical policing decisions, which means algorithms trained on it will inherit and amplify the biases embedded in those decisions.