TL;DR

West Midlands Police Chief Constable Craig Guildford has retired after it emerged that Microsoft Copilot generated a fictional match report that influenced the decision to ban Israeli fans from a football match. The incident highlights ongoing risks of AI hallucinations in operational decision-making.

A Costly AI Hallucination

The chief constable of England’s third-largest police force has retired following a significant AI-related incident. Craig Guildford, 52, stepped down on 16 January after it emerged that West Midlands Police had based part of its decision to block Maccabi Tel Aviv fans from attending an Aston Villa Europa League match on reports of disruption at a match that never happened.

The fictional content described a non-existent match between Maccabi Tel Aviv and West Ham—a game that was entirely fabricated by Microsoft Copilot.

The Cover-Up Compounds the Crisis

What made the situation particularly damaging was Guildford’s initial denial. On 6 January, he told MPs on the House of Commons home affairs committee that officers had found the material through a Google search and explicitly stated: “We do not use AI.”

Six days later, Guildford wrote to the committee chair to apologise, acknowledging he had “become aware that the erroneous result concerning the West Ham v Maccabi Tel Aviv match arose as result of a use of Microsoft Copilot.”

Home Secretary Shabana Mahmood subsequently expressed she had no confidence in Guildford.

A Pattern of AI Fabrication

This incident joins a growing list of AI hallucination cases affecting serious institutions. Lawyers in both the US and UK have cited made-up cases generated by AI tools. In October 2025, consultancy Deloitte refunded A$440,000 to the Australian government after producing a report with fabricated references and footnotes.

Looking Forward

For organisations using AI tools, this case reinforces critical lessons: AI outputs must be verified against authoritative sources, staff need training on AI limitations, and honest disclosure when AI is involved in decision-making is essential. The reputational and operational costs of unverified AI content can be severe—particularly when compounded by attempts to deny AI involvement.