Inside This Report

• 83% of police agencies now use AI— but the rules governing it are still being written after the fact.
• 44% of agencies have given officers zero AI training, even as personnel use the tools on personal devices, with or without approval.
• No U.S. agency has a proven, field-wide accountability framework for AI — while the UK has already built one.

By Samuel López | USA Herald

Artificial intelligence has already walked through the front door of American law enforcement. The problem, according to a sweeping new report from the National Policing Institute (NPI), is that nobody bothered to build the guardrails first.

The report, The AI Adoption Strategy Gap in Policing: An Unmet Need with Real-World Consequences, released August 11, lands as both a warning and a wake-up call. It confirms what many insiders have suspected for months: police agencies across the country are deploying AI tools at a pace that has sharply outrun their ability to govern, audit, or even train officers on them.

The findings are drawn from NPI's April 2026 Executive and Leaders Roundtable in San Francisco, hosted by Microsoft, which pulled together police chiefs, sheriffs, senior command staff, crime analysts, and technology strategy officers from more than a dozen agencies — from major metro departments to small-city and regional sheriffs' offices.

The numbers tell an uncomfortable story. Every single agency represented at the roundtable reported some form of AI presence in its operations, and 83% had formally deployed at least one AI tool. Yet 44% had provided no AI-specific training to any personnel — meaning officers and deputies are, in many cases, teaching themselves how to use these systems in the field, sometimes on personal devices, sometimes without their department even knowing.

Governance structures are only marginally further along. Roughly 61% of agencies said they'd designated a person or committee to oversee AI use. But according to the report, that designation is largely symbolic without answers to the questions that actually matter: Who signs off on a specific use case? What does the audit trail look like when an algorithm gets it wrong? Who's accountable when it does?

One roundtable participant, speaking under Chatham House rules that shield individual agencies and officials from attribution, offered a blunt assessment of the risk: leadership needs to act now, because rank-and-file officers are already using AI — whether or not it's sanctioned, tracked, or approved — and it's only a matter of time before that catches up with them.

James Burch, President of the National Policing Institute, was careful to frame the report not as a call for restriction, but as a call for preparation. The gap isn't a lack of rules against using AI — it's the absence of a responsible framework for adopting it well. That means building out governance and accountability processes, training the workforce, and bringing communities into the conversation early, rather than after a controversy forces the issue.

The report argues the field isn't actually short on guidance about what individual AI tools can do — vendors and decision frameworks for evaluating specific products are proliferating. What's missing is a broader strategy: the organizational readiness — data infrastructure, oversight, workforce literacy, and public trust — that needs to exist before a tool goes live, not after.

NPI points to a handful of models the rest of the field could learn from. Georgia's state-run Innovation Lab requires ethics training before any AI tool reaches production. The United Kingdom has committed $95 million to a new national PoliceAI center testing and scaling AI tools across all 43 police forces in England and Wales. And the San Diego County Sheriff's Office spent more than a decade building in-house AI capability instead of leaning entirely on outside vendors — producing tools like a deputy-built interview transcription system with projected annual operating costs under $100,000.

The most pointed comparison in the report, though, is structural. The UK's National Police Chiefs' Council has published a governing covenant built on six principles every force must follow: lawful, transparent, explainable, responsible, accountable, and robust. American law enforcement, the report states plainly, has no equivalent — no agency has developed a proven, replicable accountability framework that could serve as a national standard.

The report doesn't pretend the fix is simple. It calls for a coordinated national strategy — one that treats AI governance as infrastructure, not an afterthought bolted on once something goes wrong. For departments large and small, the message is the same: officers are already using these tools. The only open question is whether leadership catches up before the first major failure forces the issue into public view.

The full report from the National Policing Institute is available on the organization's website.

About the Author

Samuel López is a Senior Legal Analyst, investigative journalist, and legal researcher for USA Herald with more than two decades of experience dissecting high-stakes litigation, regulatory battles, and the legal frameworks that shape emerging technology. His work has focused extensively on the intersection of law enforcement, civil liberties, and institutional accountability — territory that puts him squarely at the center of stories like this one, where legal exposure, public policy, and rapidly evolving technology collide. López has spent years tracking how government agencies adopt new tools faster than the legal and regulatory systems built to oversee them, giving him a rare, informed vantage point on the risks — and the stakes — of AI's arrival in policing.

Stay Ahead of the Story

Get the investigative reporting, legal analysis, and inside-the-room coverage other outlets miss. Subscribe to the USA Herald newsletter for exclusive access to the stories shaping law, technology, and national security before they break everywhere else.

👉 Follow us on X: https://x.com/RealUSAHerald