AIPolicyTracker
Menu

AI incident ·

West Midlands Police Reportedly Relied on Erroneous Copilot-Generated Intelligence in Maccabi Tel Aviv Away-Fan Ban Decision

2 news reports Synced from source · record last edited 4 Sep 2026

In brief

An AI system built by Microsoft and deployed by West Midlands Police, National security and intelligence stakeholders and 1 other allegedly harmed Maccabi Tel Aviv supporters, Maccabi Tel Aviv F.C. and 2 others.

Risk domain
Misinformation False or misleading information
Occurred
Coverage
2 reportsJan 2026

What happened

West Midlands Police reportedly included inaccurate intelligence purportedly generated using Microsoft Copilot in materials used to justify banning Maccabi Tel Aviv supporters from attending a November 2025 Europa League match against Aston Villa. The reported Copilot-linked error, which referred to a match that had not taken place, was later acknowledged by Chief Constable Craig Guildford.

Laws that address this harm

Policy angle: Classified under Misinformation (False or misleading information) in the MIT AI Risk Repository taxonomy; 5 recorded instruments address this use case.

Matched from the record's risk domain and country to the instruments recorded here. A reviewer can correct the match in the repository (data/external/incident_overrides.yaml).

News reports (2)

Titles link to the original publisher; report text is not reproduced here.

  1. West Midlands police chief apologises after AI error used to justify Maccabi Tel Aviv ban
    theguardian.com · Neha Gohil
  2. Inspection of police forces’ contributions to safety advisory groups: West Midlands Police (accessible)
    gov.uk · Andy Cooke

Who was involved

Alleged developer
Microsoft
Alleged harmed party
Maccabi Tel Aviv supporters Maccabi Tel Aviv F.C. Jewish community in Birmingham and the West Midlands Epistemic integrity

AI systems implicated

CopilotAI-enabled decision support systems

Classification (MIT AI Risk Repository taxonomy)

Risk domain
Misinformation
Causal entity
AI
Intent
Unintentional
Timing
Post-deployment
Harm level
—
Sectors
—
Countries
—

Risk entries describing this failure mode

Entries from the MIT AI Risk Repository coded to subdomain 3.1.

  • Pursuing Consistent Context

    "LLMs have been demonstrated to pursue consistent context [129]–[132], which may lead to erroneous generation when the prefixes contain false information. Typical examples include sycophancy [129], [130], false demonstra...

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Defective Decoding Process

    In general, LLMs employ the Transformer architecture [32] and generate content in an autoregressive manner, where the prediction of the next token is conditioned on the previously generated token sequence. Such a scheme...

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Noisy Training Data

    "Another important source of hallucinations is the noise in training data, which introduces errors in the knowledge stored in model parameters [111]–[113]. Generally, the training data inherently harbors misinformation....

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Hallucinations

    "LLMs generate nonsensical, untruthful, and factual incorrect content"

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Faithfulness Errors

    "The LLM-generated content could contain inaccurate information" which is is not true to the source material or input used

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Factuality Errors

    "The LLM-generated content could contain inaccurate information" which is factually incorrect

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Untruthful Content

    "The LLM-generated content could contain inaccurate information"

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

  • Knowledge Gaps

    "Since the training corpora of LLMs can not contain all possible world knowledge [114]–[119], and it is challenging for LLMs to grasp the long-tail knowledge within their training data [120], [121], LLMs inherently posse...

    Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems (Cui2024)

Linked by editors or by text similarity in the source dataset.

Incidents in the same risk subdomain

All incidents in this subdomain

Source record: incident #1400 on the AI Incident Database · all 2 reports