Police told to pause AI use in court statements

TL;DR:

  • Several England and Wales forces have been told to stop using AI to prepare court statements until the tools are properly assessed.
  • Alex Murray, head of the new Police.AI centre, said justice-system AI must meet a standard “beyond reasonable doubt”.
  • The intervention follows a case where AI-generated material fabricated a football fixture in a police dossier.

A number of UK police forces have been ordered to halt their use of commercial AI tools for court statements and other criminal-justice tasks, amid concern that inaccurate outputs could contaminate legal proceedings. The instruction came from Alex Murray, head of the recently established Police.AI centre, who said he had intervened where forces deployed tools before proper checks.

Safeguards before scale

“Pause that. We need to slow it down a bit,” Murray recounted telling forces, insisting that any technology used in the justice system must meet a standard of accuracy “beyond reasonable doubt”. He flagged disclosure schedules — the records of evidence handed to the defence — as needing particular caution, requiring forces to show how a system was trained and deployed.

The warning is grounded in a concrete failure. Last year West Midlands Police relied on Microsoft Copilot output that invented a past match involving Maccabi Tel Aviv, used in a dossier supporting a proposed supporter ban. The episode crystallised fears about “hallucinations”, where systems produce convincing falsehoods. Police.AI was set up this year as part of efficiency reforms; Home Secretary Shabana Mahmood has said the £115m, three-year initiative could free up time equivalent to 3,000 extra officers across the roughly 145,000 serving.

Looking forward

Murray is not anti-AI — he pointed to gains in scanning CCTV and triaging child-abuse imagery so officers view less harmful material. But the intervention marks a notable instance of a UK public body actively reining in deployment rather than encouraging it, a counterpoint to the government’s wider adoption drive. For organisations in regulated sectors, the lesson generalises: where AI output carries legal or safety weight, demonstrable assurance must precede rollout. Expect the “beyond reasonable doubt” bar to shape how justice-system AI is procured and audited.

Share this article