Microsoft Copilot Fabricates Football Match in Police Gaffe

The West Midlands Police now owns a rather public lesson in digital literacy. An intelligence report, a document meant to inform decisions, to guide actions, to safeguard public order, was fatally undermined by a bot that hallucinated a football match. Not a misunderstanding of data, not a misreading of a tweet: a full fabrication.

Copilot told a lie, and the police printed it as fact. Guildford’s about-face, first blaming social media scraping, then pointing at Microsoft’s own tool, is telling. It reveals an institution still fumbling with the basic question: when does an AI assistant become a liability?

The company’s boilerplate disclaimer, “Copilot may make mistakes,” is a shrug wrapped in legalese. But here, the mistake wasn’t trivial. It embroiled a police force in a false narrative about a high-risk match, one already tangled in real tensions and real bans.

So where does the responsibility land? On the chief constable who trusted a probabilistic text generator without verification? On the tool that confidently produced falsehoods?

Or on a system that pushes AI into mission-critical workflows before the sharp edges are sanded down? The answer, uncomfortably, is all of the above. Until tools stop “making stuff up” and humans stop taking them at their word, this won’t be the last blunder.

The West Ham–Maccabi Tel Aviv match never happened. But the cautionary tale it leaves behind is very, very real.