195 incidents on record · 2026 Headlights Incident reports by Ellie Harris · Melbourne
11 new this week Library last updated 12 September 2026
← The incident library
HD-INC-179
Policing · United Kingdom · 2025 · Inaccurate intelligence, including an AI sourced claim not verified or removed, used in a public order decision

Police intelligence used to support an away fan ban contained eight reported errors, including a Copilot result about a match that never happened

By Ellie Harris · Filed Match 6 November 2025; AI use denied to a parliamentary committee 6 January 2026 and attributed to Microsoft Copilot in a letter dated 12 January 2026

Alleged: West Midlands Police developed or deployed the AI system implicated in this incident. Details are drawn from public reports; parties are presumed innocent of any wrongdoing not established by an official finding.

Police intelligence used to support an away fan ban contained eight reported errors, including a Copilot result about a match that never happened

What happened

It was reported that West Midlands Police supplied intelligence that underpinned a decision to ban Maccabi Tel Aviv supporters from attending a Europa League match against Aston Villa on 6 November 2025.

It was reported that eight errors were later identified in the police material. One overstated Maccabi supporters’ involvement in disorder in Amsterdam. Another referred to a match between West Ham United and Maccabi Tel Aviv that had never taken place. That error was later attributed to a Microsoft Copilot search.

It was reported that on 6 January 2026, the chief constable told the House of Commons Home Affairs Committee that the force had not used AI. Six days later, he wrote to the committee saying the incorrect information about the West Ham match had come from a Microsoft Copilot search. He retired on 16 January.

What an auditable version would have shown

The Copilot result is the clearest error because the underlying fact was easy to check. The two clubs had never played each other.

But this was wider than one AI sourced result. Eight errors were reported in the police material, and the IOPC is now examining how the intelligence was gathered, checked and handled.

It was reported that the Copilot connection became known while material was being prepared for an inspection. The sources reviewed do not show that the force already had a record identifying which parts of the intelligence had involved AI.

A record for each material claim could have shown where it came from, whether AI was involved, who checked it and what that check found.

Where the gap was

It was reported that the IOPC is investigating how the briefings and intelligence report were prepared, how information was obtained, what checks were made and how records were kept. Those are questions under investigation, not findings.

A VerificationGate is designed to check a factual claim against a trusted source before it is relied on. In this case, whether two football clubs had ever played each other could be checked against the fixture record. A ConductRecord keeps the source, any AI involvement, the check and the wording that eventually appears in the intelligence product.

Neither decides what public order advice police should give.

What governance should have looked like

It was reported that the IOPC’s update of 26 August 2026 said it had examined more than 600 pieces of evidence, and that it is investigating the former chief constable, another senior officer and a staff member for potential gross misconduct, with two other senior officers being investigated for potential misconduct.

That reporting also records the IOPC as alleging that the staff member involved in researching and compiling intelligence sourced inaccurate information from AI searches and took no subsequent action to help ensure it was excluded from the briefings. These remain allegations under investigation. The IOPC has made no finding that misconduct occurred, and serving a notice does not mean disciplinary proceedings will follow.

Where an organisation relies on intelligence to support a consequential decision, best practice would be to keep a record for every material factual claim showing where it came from, whether AI contributed to it, who checked it and what the check found before that information was relied on.

Failure Pattern: it was reported that police intelligence underpinning a restriction on supporters contained eight errors, including a reference to a football match that had never happened. The incorrect match was later attributed to a Microsoft Copilot search.

Governance Principle: where an organisation relies on intelligence to support a consequential decision, it should be able to show where each material claim came from, whether AI contributed to it and how it was checked before being relied on.

The reference implementation of VerificationGate and ConductRecord is open source. It lives at github.com/saffronandindia/headlights-oss, Apache 2.0 licensed and free to install. The repository is public now.

Sources

The mailing list

Fresh incident reports every week. One email to match.

We add new incidents to the library regularly, and send a single short email each week with what's new. The library stays free and open; this is just how you keep up with it.

No tracking. Unsubscribe in one click.

The record

An auditable system would have produced a signed, tamper-evident record the moment this happened: what the system did, the version that did it, the basis it acted on, and the action taken, and West Midlands Police could have produced it on demand.

This is the record the system as deployed did not produce in a signed, auditable form.

What this teaches
Capture what happened when it happens
What the system did, the version that did it, the basis it acted on, and the action taken, recorded at the moment, not reconstructed after.
Sign it, so no one has to trust the record-keeper
A tamper-evident entry. Edit it later and the signature breaks. The record does not ask for the benefit of the doubt.
Make it verifiable by anyone
A court, a regulator, a customer's lawyer can check the record themselves, without taking the company, or us, at our word.

Headlights summarises publicly reported AI incidents. All summaries are independently written, attributed to their original sources, and intended for research and educational purposes. Allegations are identified as such until established through official findings.

Last reviewed August 2026. This report is based on the sources listed above and reflects information available at the time of review; later developments may not be captured. Where a person is described as charged with or alleged to have done something, that allegation is unproven unless a conviction or a court or regulatory finding is stated. Headlights publishes journalism and commentary, not legal advice.

Want to write back?

Direct to my inbox.

ellie@useheadlights.com →