Police used fabricated information created by AI to ban Israeli fans from match; chief lied about where it came from
West Midlands Police in the United Kingdom is under scrutiny after chief constable Craig Guildford admitted that artificial intelligence-generated information helped justify a decision to bar Israeli club Maccabi Tel Aviv’s supporters from a Europa League match against Aston Villa in Birmingham.
For senior HR leaders, the case is a reminder that when organisations lean on generative AI for safety, risk or people decisions, unverified “hallucinations” can quickly become real-world liability.
The match in November was classed as high risk by West Midlands Police. In advice to a local Safety Advisory Group, the force cited past trouble involving Maccabi Tel Aviv fans, including violent clashes and hate crime offences supposedly linked to a match against West Ham United – a game that never took place.
Erroneous AI information initially blamed on Google search
Guildford has now told the UK Parliament’s Home Affairs Select Committee that material he and a senior colleague had previously blamed on a Google search was in fact produced by Microsoft Copilot, according to multipole media outlets.
He has apologised to MPs for providing “erroneous” evidence, having earlier insisted before the Home Affairs Select Committee that AI tools were not used to prepare intelligence reports.
After the controversy grew, Home Secretary Shabana Mahmood ordered an inspection by His Majesty’s Inspectorate of Constabulary into how intelligence supporting the fan ban was gathered and handled. A Home Office spokesperson told media that the report examines the recommendation by West Midlands Police to prevent fans from travelling to the match.
Mahmood later told MPs she had lost confidence in the chief constable, describing the review as “damning” and saying police overstated the threat posed by the Israeli fans while understating the risk they faced if they travelled to the area, according to the BCC. She said misleading communications extended to Guildford’s own evidence to Parliament and noted that one of his officers blamed incorrect evidence on an “AI hallucination.”
UK media report that the force is facing calls for the chief constable to step down, and that the regional police and crime commissioner will review decisions made on the ban.
Why this matters for employers using AI
For HR leaders, the episode is a case study in how not to use generative AI.
First, AI hallucinations are a governance problem, not just a technical quirk. Generative systems are known to produce material that is plausible but false. In this case, an invented football match was treated as credible intelligence and helped shape a high-profile public safety decision.
In an employment setting, similar hallucinations could influence hiring, promotion, discipline or termination if AI is used to summarise investigation files, background checks, disciplinary histories or safety reports.
Second, human review must be explicit and accountable. According to media reports in this case, AI-generated material ended up in a formal submission to a safety advisory group and in evidence to MPs without sufficient verification.
Many employers are embedding tools such as Microsoft Copilot into office software, HR platforms and case management systems. Without clear labels showing when content is AI-generated, and without mandatory verification steps, fabricated or biased material can easily blend into official records.
Third, AI use in sensitive contexts demands heightened scrutiny. The West Midlands case involved public safety, community tensions and international politics. HR leaders face comparable sensitivities in workplace violence and harassment investigations, labour disputes, discrimination or hate speech complaints, and major restructurings. In such settings, even a small AI-driven error can damage trust and be read as evidence of systemic bias or negligence.
Lack of transparency about AI use erodes trust
Fourth, transparency about AI use is essential. Senior officers initially told MPs that incorrect intelligence came from a Google search. Only later did it emerge that AI tools had been used, raising concerns about candour with Parliament and the public.
Employers that hide or downplay AI use in HR decisions risk similar backlash from regulators, tribunals, unions and employees.
Finally, leadership accountability cannot be delegated to software. Mahmood’s conclusion that she no longer had confidence in the chief constable underlines that leaders remain personally responsible for how AI is deployed, even when tools come from large, trusted vendors.
