UK police chief resigns over AI intel embarrassment, blames 'political and media frenzy' quite than his power's failure to fact-check Copilot
West Midlands Police chief constable Craig Guildford, the top of the UK police power that admitted earlier this week that it had used defective data from Microsoft’s Copilot AI assistant in its controversial resolution to ban Israeli soccer followers from a November 2025 Europa league match, has introduced his resignation (through the BBC). His retirement follows days of criticism over the scandal, with UK house secretary Shabana Mahmood and Downing Avenue each expressing that they’d “misplaced confidence” within the chief constable.
Moderately than his power’s failure to fact-check a notoriously unreliable expertise, Guildford mentioned in a press release saying his retirement that “the political and media frenzy” was liable for his resolution to resign.
“The political and media frenzy round myself and my place has turn out to be detrimental to all the nice work undertaken by my officers and employees in serving communities throughout the West Midlands,” Guildford mentioned. “I’ve rigorously thought of my place and concluded that retirement is in the perfect pursuits of the organisation, myself and my household. It has been the honour of my profession serving because the Chief Constable of West Midlands Police.”
In his assertion, Guildford didn’t apologize for the West Midlands Police’s dealing with of the Aston Villa v Maccabi Tel Aviv match, or for its inclusion of factually inaccurate data within the previous intelligence report.
In November 2025, the Birmingham Security Advisory Group—led by West Midlands Police and Birmingham Metropolis Council—made the contentious resolution to ban Israeli followers from attending an upcoming Europa League match between Aston Villa and Maccabi Tel Aviv, fearing a repeat of November 2024’s violent clashes over Gaza struggle tensions with Israeli soccer followers in Amsterdam.
That call was knowledgeable by an intelligence report submitted by the West Midlands Police, which—along with that includes interpretations of the 2024 Amsterdam fan violence which have since been questioned by Dutch authorities—referenced an earlier West Ham v Maccabi Tel Aviv match in 2023. That West Ham v Maccabi Tel Aviv match, nonetheless, by no means occurred, which is the precise sort of error you do not need in an intelligence report.
In a letter to the House Affairs Committee earlier this week, Guildford admitted the West Ham match was fabricated by Microsoft’s Copilot AI assistant, which the WMP had consulted and didn’t fact-check whereas compiling its report.
“On Friday afternoon I turned conscious that the inaccurate outcome regarding the West Ham v Maccabi Tel Aviv match arose as results of a use of Microsoft Copilot,” Guildford mentioned. “Each ACC O’Hara and I had, up till Friday afternoon, understood that the West Ham match had solely been recognized by using Google.”
In December, Guildford had been questioned in a House Affairs Committee listening to investigating the dealing with of the Aston Villa v Maccabi Tel Aviv match. When requested by a minister of parliament whether or not the WMP “did an AI search, received one thing about West Ham, and simply whacked it into” the report, Guildford mentioned “No, by no means,” claiming he as an alternative believed the nonexistent soccer match to have emerged from the police power’s “search by social media.”
In his letter, Guildford acknowledged that the WMP had certainly simply whacked the outcomes of an AI search into its intelligence report, and supplied his “profound apology to the Committee for this error.”
Like different LLM-driven assistants, Copilot contains a disclaimer that it “could make errors,” which you would possibly hope would encourage the personnel of safety, legislation enforcement, and protection organizations who’re using LLM applied sciences to train due warning. However when these applied sciences, regardless of their notorious unreliability, are marketed as one-button replacements for human information and problem-solving, it is onerous to think about a lot of these oversights as being something however inevitable.
In November, Microsoft AI CEO Mustafa Suleyman mentioned it is “mindblowing” that folks aren’t extra impressed with generative AI. Mustafa, for those who’re listening: This is the reason. Fortunately, it is not like we’re setting ourselves up for even greater stakes AI embarrassments.
In different information, US secretary of protection Pete Hegseth introduced this week that the Pentagon will start integrating X’s Grok AI into navy networks. Sure, the identical LLM platform drawing rising outrage for flooding X with generated pornographic imagery of nonconsenting folks, together with minors. Absolutely nothing may go unsuitable with that, both.