Generative AI

Generative AI produces new text rather than matching or scoring, and it is reaching policing mostly as general office software, such as Microsoft 365 Copilot, rather than as tools bought and tested for police work.

Guide

What it is

  • General assistants tools such as Copilot that summarise documents, draft correspondence and condense case material.
  • Report drafting from footage products sold to US police that draft an incident narrative from body-worn camera audio.
  • Hallucination language models can produce fluent, confident text that is false, with no sign that it was invented.

How it is used

Copilot reached English forces through the Police Digital Service's shared cloud platform in 2025, and a Sky News survey found at least 21 forces kept using it after West Midlands Police withdrew it. The record tracks which forces use which tools, incidents where output went wrong, and the rules forces set for themselves.

Key tensions

The West Midlands incident

In November 2025 Copilot invented a disorderly match that fed into an intelligence assessment and a decision to bar away fans. The chief constable first told MPs the force did not use AI, corrected this, and retired.

Confident errors in trusted documents

Police documents are trusted because of where they come from. A tool that adds unattributed errors to intelligence reports, case summaries or statements undermines the part of the process nobody expects to question.

Disclosure, and 43 sets of rules

It is unsettled whether AI drafting must be disclosed in criminal cases, and there is no dedicated UK law. After West Midlands, only eight surveyed forces confirmed Copilot could not be used in investigations.

View all generative AI records Latest coverage

In depth

The full guide, about 6 minutes to read.

What generative AI means in a policing context

Most of the artificial intelligence covered elsewhere on this site does one narrow thing very quickly: match a face against a watchlist, read a number plate, score a risk. Generative AI is different in kind. It produces new text, and it produces it fluently, which turns out to matter a great deal when the output is going into a police record.

The systems now reaching policing are, for the most part, not policing products at all. Microsoft 365 Copilot is a general-purpose assistant built into standard office software, sold to organisations of every type. Its arrival in English policing came through the Police Digital Service's National Police Capabilities Environment, an Azure-based shared platform launched in 2025 to give forces common access to cloud tools. Copilot came with it, in the way that a spellchecker comes with a word processor, rather than as a considered policing procurement in its own right.

That distinction matters more than it might sound. A facial recognition system bought for policing arrives with accuracy testing, a threshold setting, an oversight framework and a body of case law. A general office assistant arrives with none of those, because nothing about its design anticipated that its output might end up in an intelligence assessment or a case file.

Report writing, the first mainstream use

The administrative burden in policing is real and well documented. Officers spend a substantial proportion of their time writing: incident reports, statements, handover notes, case file summaries. Any technology that credibly reduces that burden has an obvious appeal, and it is the reason generative AI has spread through forces faster than more visible technologies with far smaller footprints.

A Sky News survey found that at least 21 English police forces continued using Copilot after West Midlands Police withdrew it, and that some forces were using the free consumer chat product rather than the enterprise version designed for workplace governance, which is a materially different proposition in a policing context. Typical described uses are summarising long documents, drafting routine correspondence, and condensing case material. Separately, some US vendors sell tools built specifically for policing that draft an incident narrative directly from body-worn camera audio, which is a materially more consequential application than summarising an email thread.

The efficiency argument for all of this is straightforward and largely uncontested. The dispute is about what happens at the edges: what the system does when it does not know something, and whether anyone downstream can tell.

The West Midlands incident

In November 2025, West Midlands Police were preparing an intelligence assessment ahead of a Europa League fixture between Aston Villa and Maccabi Tel Aviv. Copilot was used during that preparation and produced a reference to a match involving disorder that had never taken place. The fabricated detail fed into the deliberations of a local Safety Advisory Group, which decided to bar away fans from attending.

The subsequent handling is as instructive as the error. Chief Constable Craig Guildford told the House of Commons Home Affairs Committee on 6 January 2026 that officers had not used AI, saying in evidence: "We do not use AI." He corrected that account in a letter to the committee chair dated 12 January, writing that the erroneous result had arisen from a use of Microsoft Copilot. He retired on 16 January, with the West Midlands Police and Crime Commissioner stating he had already begun the process that would have allowed him to dismiss Guildford. Acting Chief Constable Scott Green told a public meeting on 27 January that he had turned off Copilot access across the force on his first day in office, and would not restore it until the investigation concluded and broader policy and guidelines were in place. The Independent Office for Police Conduct and the force's own professional standards department are both investigating.

The Home Affairs Committee published its findings in February 2026, concluding that "proper due diligence was not applied" and noting that Copilot had produced a second inaccuracy, this time about disorder at a 2024 Maccabi fixture in Amsterdam. The West Midlands Police and Crime Commissioner described "significant concerns, shortcomings, and failures" in how AI use was governed during the operation.

What makes this case genuinely significant is not that a language model produced something false, which is a well-understood property of the technology. It is that the false detail travelled: from a drafting tool, into an intelligence assessment, into a multi-agency decision-making forum, and out into a decision that affected several thousand people, without anyone in that chain identifying it as unverified. Full sourcing for this account is on the deployment tracker, and the system itself is covered on the Microsoft 365 Copilot technology page.

Why hallucination is a different kind of risk here

Large language models generate text by predicting plausible continuations. They do not have a mechanism for distinguishing between a fact they have encoded and a fluent invention, which is why fabricated output tends to arrive with exactly the same confident register as accurate output. In most settings that produces embarrassment. In policing it can produce a wrongful decision.

The specific difficulty is that policing documents are trusted downstream precisely because of where they come from. An intelligence assessment carries weight in a Safety Advisory Group, a case summary carries weight with a prosecutor, a report carries weight in court. That trust is institutional rather than textual: nobody reads a force intelligence document expecting to have to fact-check its internal references. Generative AI inserts a source of confident, unattributed error into exactly the part of the process that is designed not to be questioned.

This is a distinct problem from the accuracy debates around facial recognition or risk scoring, where the failure mode is a measurable false-positive rate that can at least be tested, published and argued about. There is no equivalent published error rate for a general-purpose assistant used to draft an intelligence report, because the task has no defined ground truth to measure against.

Evidential and disclosure problems

Criminal proceedings run on disclosure. The defence is entitled to understand how material against a defendant was produced, and the prosecution carries continuing obligations to disclose material that might undermine its case or assist the defence. Generative AI complicates that in ways the law has not yet fully worked through.

If an officer's statement was drafted with AI assistance from body-worn footage, several questions follow that currently have no settled answer. Is the fact of AI assistance itself disclosable? Does the original audio become the primary record with the statement as a derived document? If the model summarised or paraphrased, what happened to the material it left out, and would anyone know it had been left out? A model does not flag its own omissions.

There is also the question of what a defendant can meaningfully challenge. Where a risk-assessment score is disputed, there is at least a model, a set of inputs and a methodology to interrogate, however opaque. Where a narrative was drafted by a general-purpose assistant that no longer exists in the same state, that has been updated repeatedly since, and whose output is not reproducible, the thing being challenged is much harder to pin down.

Where the rules currently sit

There is no dedicated statutory framework for police use of generative AI in the UK. The governing rules are the general ones: data protection law where personal data is processed, human rights law where a decision engages a protected right, disclosure obligations in criminal proceedings, and force-level policy.

Force-level policy appears to vary considerably. Following the West Midlands incident, only eight of the forces surveyed confirmed that Copilot could not be used in investigations, which implies that the others either permit some investigative use or have not drawn the line explicitly. That variation is itself notable: a technology deployed through a shared national platform is being governed by forty-three separate local judgements about what it may be used for.

The Home Office's Algorithmic Transparency Recording Standard offers a voluntary route for public bodies to publish structured records of algorithmic tools in use, which would in principle cover this. Uptake among police forces has so far been limited, which means the clearest public picture of how generative AI is being used in policing has come from journalism and parliamentary questioning rather than from routine disclosure.

Records on the tracker

Open the full tracker

48 records on generative AI, most recently checked first: 41 in use now, 1 withdrawn and 6 on the horizon. Each is one police force or public body using one system.

  1. Spain

    Fiscalía General del Estado (Spain) — AI tool drafting charging documents from police reports (drink-driving)

    Spain, national

    TrialChecked
  2. Denmark

    Anklagemyndigheden (Danish Prosecution Service) — AI assistant for legal information searches in criminal case handling

    ProposedChecked
  3. Norway2025

    Oslo politidistrikt (Oslo Police District) — AI4Interviews: speech-to-text transcription of interviews and AI-drafted reports

    Oslo

    TrialChecked
  4. Norway2025

    Vest politidistrikt (Western Police District) — AI4Interviews: speech-to-text transcription of interviews and AI-drafted reports

    Bergen and Western Norway

    TrialChecked
  5. United StatesPalantir Technologies2025

    US Immigration and Customs Enforcement — AI-enhanced tip processing (large language model summaries of web tips)

    Nationwide

    OperationalChecked
  6. United StatesDeveloped and vetted by a US Intelligence Community element outside the DEA and DOJ (not named)

    Drug Enforcement Administration — AI triage and summarisation of open-source foreign reporting

    Nationwide (Office of National Security Intelligence)

    OperationalChecked
  7. United States2024

    ICE Homeland Security Investigations — Generative AI summaries and translation of dark web data for cyber investigations

    Nationwide

    OperationalChecked
  8. United StatesPenlink, Fivecast2023

    ICE Homeland Security Investigations — Penlink and Fivecast open-source intelligence platforms with generative AI

    Nationwide

    OperationalChecked
  9. United StatesConductor

    Executive Office for United States Attorneys — Conduit AI document review platform (pilot)

    Nationwide

    TrialChecked
  10. United States

    US Coast Guard — TRITON voice-enabled boarding report app

    ProposedChecked
  11. United StatesAxon

    Fairfax County Police Department — Axon Draft One AI report drafting

    Fairfax County, Virginia

    OperationalChecked
  12. United StatesAxon2026

    Hillsborough County Sheriff's Office — Axon Draft One AI report drafting

    Hillsborough County, Florida

    OperationalChecked
  13. United StatesBuilt by the department on Microsoft Azure Government

    Florida Department of Corrections — CLEO generative AI assistant

    Florida

    OperationalChecked
  14. United States

    Connecticut Criminal Justice Information System (CJIS) — AI chatbot helping CISS users build search terms

    Connecticut

    OperationalChecked
  15. United StatesPoliceNarratives.AI

    Manchester Police Department (New Hampshire) — PoliceNarratives.AI draft police report generation

    Manchester, New Hampshire

    TrialChecked
  16. United StatesAxon

    Manchester Police Department (New Hampshire) — Axon Draft One pilot (AI-drafted incident reports from body-camera audio)

    Manchester, New Hampshire

    DiscontinuedChecked
  17. United StatesAxon

    New Haven Police Department — Axon Draft One (AI-drafted police reports from body-camera audio)

    New Haven, Connecticut

    TrialChecked
  18. United StatesAxon

    Connecticut State Police — Axon Draft One module within body-camera contract (AI-drafted police reports)

    Connecticut

    ProposedChecked
  19. United KingdomMicrosoft

    Sussex Police — Microsoft Copilot (generative AI)

    Sussex

    OperationalChecked
  20. United KingdomMicrosoft

    Surrey Police — Microsoft Copilot (generative AI)

    Surrey

    OperationalChecked
  21. United Kingdom

    Sussex Police — AI speech-to-text transcription pilot

    Sussex

    TrialChecked
  22. United KingdomGreat Wave AI (initial development)

    Warwickshire Police — AI agents auditing investigation quality (Great Wave AI)

    Warwickshire

    OperationalChecked
  23. United KingdomMicrosoft

    Metropolitan Police — Microsoft Copilot Chat (generative AI)

    London

    OperationalChecked
  24. United Kingdom

    Metropolitan Police — AI-assisted evidence translation

    London

    OperationalChecked
  25. United Kingdom

    Metropolitan Police — AI drafting of witness statements from video interviews

    London

    PausedChecked
  26. United Kingdom

    Police Service of Northern Ireland — AI voice-to-text transcription of interviews

    Northern Ireland

    OperationalChecked
  27. United KingdomMicrosoft2025

    Police Scotland — Microsoft Copilot corporate efficiency trial

    Scotland

    TrialChecked
  28. United KingdomNTT Data UK (supplier); Microsoft Azure OpenAI GPT-42025

    Crown Prosecution Service — Correspondence Drafting Tool (generative AI)

    England and Wales

    TrialChecked
  29. United KingdomMicrosoft

    Ministry of Justice — Microsoft Copilot Chat and Microsoft 365 Copilot (AI for All)

    England and Wales

    OperationalChecked
  30. United Kingdom2025

    HM Prison and Probation Service — Justice Transcribe (AI transcription for probation meetings)

    England and Wales

    OperationalChecked
  31. United Kingdom

    HM Courts and Tribunals Service — Justice Transcribe court transcript study (Crown Court)

    England and Wales

    ProposedChecked
  32. Jersey2025

    States of Jersey Police — AI transcription of witness interviews

    TrialChecked
  33. France

    Gendarmerie nationale — "Parole" speech-to-text transcription of child victim interviews

    OperationalChecked
  34. Switzerland

    Kantonspolizei Aargau — AargauGPT generative AI assistant

    Canton of Aargau

    OperationalChecked
  35. Germany

    North Rhine-Westphalia Police — Planned generative AI tools for text, translation and case-file research on a police-run AI infrastructure

    North Rhine-Westphalia

    ProposedChecked
  36. SwitzerlandDeeplink2026

    Police Région Morges — Trinity AI tool drafting incident reports

    Canton of Vaud

    OperationalChecked
  37. United KingdomMicrosoft

    Greater Manchester Police — Microsoft 365 Copilot (generative AI)

    OperationalChecked
  38. United KingdomMicrosoft

    West Yorkshire Police — Microsoft 365 Copilot (generative AI)

    OperationalChecked
  39. United StatesAxon (Draft One)2024

    Fresno Police Department — AI-assisted police report writing

    California

    OperationalChecked
  40. CanadaAxon (Draft One)2025

    Royal Canadian Mounted Police (British Columbia) — AI-assisted police report writing

    British Columbia

    TrialChecked
  41. United KingdomMicrosoft

    Cheshire Constabulary — Microsoft Copilot (generative AI)

    Cheshire

    OperationalChecked
  42. United KingdomMicrosoft

    Avon and Somerset Police — Microsoft Copilot Chat (generative AI)

    Avon and Somerset

    OperationalChecked
  43. United KingdomAnathem

    Hertfordshire Constabulary — ADA (Anathem Digital Assistant) AI witness statement drafting

    Hertfordshire

    PausedChecked
  44. United KingdomMicrosoft2025

    Derbyshire Constabulary — Microsoft Copilot (generative AI)

    Derbyshire

    OperationalChecked
  45. United KingdomDefence Holdings (Defence Technologies)2025

    Gloucestershire Constabulary — AI summaries of police interview records

    Gloucestershire

    TrialChecked
  46. United StatesAxon (Draft One)2025

    Wheat Ridge Police Department — AI-assisted police report writing

    Colorado

    OperationalChecked
  47. United StatesAxon2026

    Lakeland Police Department — AI-assisted police report writing

    Florida

    ProposedChecked
  48. United KingdomMicrosoft2025

    West Midlands Police — Microsoft 365 Copilot (generative AI)

    West Midlands

    PausedChecked

Latest coverage

All 51 stories