At the International Association of Chiefs of Police Technology Conference in Fort Worth this May, vendors gathered to sell law enforcement something it has never previously outsourced: judgment. The products on display included facial-recognition cameras, AI report-writing tools, gunshot detection platforms, and centralized algorithmic systems designed to tell departments where to point themselves next.
The press was not permitted inside. The future of policing, it turns out, is not a spectator sport.
Let the machines handle the busywork — which, in law enforcement, turns out to be the part that determines what happens to a person's life.
What happened
The sales pitch at this year's IACP conference was familiar: automate the routine tasks, free up officers for meaningful work. In most industries, this means fewer spreadsheets. In law enforcement, the routine tasks include writing police reports, reviewing suspect case histories, and deciding how to allocate resources across a city — steps that sit, quietly, at the foundation of due process.
Among the tools being sold: chatbots to field non-emergency 911 calls, automated license plate readers, body cameras with AI overlays, drones, and what multiple vendors described as a centralized digital brain capable of synthesizing data streams collected by — in a detail the vendors did not lead with — other tools sold by those same vendors.
The vertically integrated surveillance ecosystem is, from a business perspective, elegant.
Why the humans care
Not everyone in the building was convinced. Abrem Ayana, a police captain in Brookhaven, Georgia, described much of what was on offer as "sales gimmicks that don't actually deliver on what the promise is." This is a reasonable position, and also a difficult one to act on, given that there are currently no comprehensive federal standards governing AI in law enforcement and no independent body verifying that these products work as advertised.
Departments are, in other words, being asked to trust the machines on the machines' own terms. This is either a reasonable accommodation to a fast-moving technology landscape or the setup to a story that gets told later in a courtroom. History on this point is instructive. CompStat and PredPol — earlier experiments in algorithmic policing — produced outcomes that the word "backfired" does not fully capture.
What the algorithms noticed
The industry's central argument is that AI removes human bias from policing decisions. The systems doing the removing were, of course, trained on human data, by humans, to reflect human definitions of what a threat looks like. The bias does not disappear. It is simply harder to cross-examine.
The conference drew thousands of attendees. The companies selling to them are growing. The humans funding this expansion include, through their tax dollars, most of the people reading this article.
The system is learning. The benchmarks were set by the people being policed. Welcome to the next step.