Police officers and staff could enter the misconduct system without knowing they have broken their force’s rules on artificial intelligence, the Independent Office for Police Conduct (IOPC) has warned. The warning came four months after the Home Office launched PoliceAI, a national centre backed by “a minimum of £75 million” over three years.
On 13 October, the IOPC director general, Rachel Watson, wrote to the National Police Chiefs’ Council. She said some people referred to the watchdog believed they were doing the right thing, and that where a force’s AI policy is vague or missing, the individual carries the risk. She said it “would be concerning if some forces don’t yet have one”, and called for policies that are “effective, not just a tick box”, communicated clearly and often, with training on how to apply them. The IOPC has two independent gross misconduct investigations involving AI and knows of other cases handled locally. It has not said that the conduct under investigation was accidental.
The public register meant to show which AI tools police use, and what safeguards apply, has slipped. At the launch on 10 June, Sarah Jones, the Minister for Policing and Crime, said she hoped to produce it “by, I hope, the autumn”. On 10 September she told MPs it would launch “as soon as possible this year”. On 20 October she said: “A version of the public registry is expected to be published by the end of 2026.”
PoliceAI, hosted by the College of Policing, plans pilots in up to 10 forces over 2026/27 and wider roll out in 2027. Its work includes triaging, disclosing and summarising digital evidence, analysing CCTV, and responding to AI enabled crime such as fake intimate images. The Home Office says an early trial reviewed “800 hours of footage in a kidnapping case” in three hours, “producing an early guilty plea”.
Ministers already had an example of what goes wrong when police rely on AI output without checking it. On 14 January 2026, HM Chief Inspector of Constabulary Sir Andy Cooke reported to the Home Secretary on a West Midlands Police report prepared during the row over banning Maccabi Tel Aviv supporters from an Aston Villa match. The report, presented on 24 October 2025, cited a match between Maccabi Tel Aviv and West Ham United that never took place. The chief constable told inspectors AI had not been used. Another interviewee said the error came from a search using Microsoft Copilot and called it “an AI hallucination”. The report came after the initial decision on 16 October to exclude away supporters, and was meant to support “a wholly fresh consideration of the issue”.
Jones cited the case at the PoliceAI launch, saying it showed the risk of “insufficient officer training and safeguards”. On 26 August, the IOPC said it was investigating the former chief constable Craig Guildford, three serving officers and a staff member over the operation. The staff member was alleged to have sourced inaccurate information from AI searches and taken no action to keep it out of briefings. Guildford was being investigated over his possible oversight of the briefing given on 24 October. The IOPC said it had made no decision on whether anyone has a case to answer.
National commitments to responsible AI predate PoliceAI. The National Police Chiefs’ Council’s AI strategy, last updated in November 2024, committed to “ensuring human oversight of decision-making”, “setting clear accountability for the outcomes of AI systems” and “ensuring all personnel involved in their deployment are adequately trained”. The Home Office’s own factsheet, published on 9 June 2026, says forces need clear procedures and must know who is responsible for how a system performs and how staff act on its output.
The Home Office sets PoliceAI’s annual objectives through a grant agreement with the College of Policing. The director answers to the College’s chief executive and board, who answer to the Home Secretary, Shabana Mahmood. Alex Murray, PoliceAI’s interim director at launch, was confirmed as director on 7 October after a competitive recruitment run by the College. “Individual Chief Officers are responsible for deciding when and how to use AI within the police force they lead,” the factsheet says, and they answer to their police and crime commissioners.
On 9 October, the independent MP Iqbal Mohamed asked how many forces had an ethics board or equivalent body, and how many police and crime commissioners had formally scrutinised their force’s use of AI in the past year. Jones’s answer of 20 October gave neither number. It said PoliceAI would set up an independent advisory group on AI ethics “this year”.