International research · Government & public sector
Government AI Hallucination Tracker
Public documents. Public decisions. Questions of reliability.
The Government AI Hallucination Tracker brings together public sources discussing suspected or confirmed AI-generated errors, fabricated/false citations and related inaccuracies in government and public-sector activity worldwide.
The focus is what each source records about the material, the possible role of AI and any response. A report of a concern, an admission and an established finding are different things.
This tracker forms part of Matthew Lee’s lecturing and teaching in law and writing and editing legal articles and reports. It is communicated solely in connection with promoting or advertising his practice at Doughty Street Chambers. Nothing here is legal advice. No Public or Direct Access. All instructions are through his clerks.
The public research record
Search the Government AI Hallucination Tracker
Search by country, document, action or words in the recorded issue. The public table keeps four columns: Date, Country, Document/Action and Issues. Where a document or action is linked, open it to read the source and its context.
The date is the date recorded for that entry. The Issues column is a research summary, not a finding made by this website. Check the source for the precise allegation, acknowledgement, correction or finding, and for any later developments.
| Date | Country | Document/Action | Issues |
|---|
This is a developing, selective collection. Its entries do not measure how often government AI hallucinations occur, and the absence of an entry does not establish that no issue arose.
What belongs in this collection
Government AI hallucinations and related inaccuracies
Policy, reports and legislation
Public sources discussing unreliable references, statistics, quotations or claims in policy papers, consultations, official reports and legislative materials.
Decisions and official evidence
Reported concerns about information used in public administration, decision-making, official submissions or evidence connected with a public body.
Public-facing information
Relevant discussion of inaccurate or disputed information in official communications, guidance and AI-assisted public services.
The Government AI Hallucination Tracker considers the reliability of material used or communicated in a public-sector context. A report prepared by an outside contractor may also be relevant where it was commissioned, published or relied on by a public body. That connection does not, by itself, establish who produced an error or who was responsible for it.
The research includes suspected and confirmed government AI hallucinations alongside related citation or factual errors for which AI involvement is disputed, unclear or absent. The source’s wording and the limits of its evidence remain central.
Evidence and context
How to read the Government AI Hallucination Tracker
What does an entry establish?
Only what the underlying source supports. A suspicion that AI was used is different from an admission; an identified inaccuracy is different from proof of its origin. A source may record someone else’s allegation without accepting it. Inclusion must not be read as adding a finding or criticism beyond the source.
Does an error mean AI was involved?
No. False references, inaccurate statistics and misleading descriptions can arise without AI. The label “AI hallucination” is not treated here as an explanation for every mistake. Relevant records may expressly leave the connection with AI unresolved.
Does inclusion mean a government decision was unlawful?
No. This collection does not determine legal responsibility or the validity of a policy, decision or document. It brings public material together for research and legal commentary. It is not advice about challenging a decision or taking any other action.
Are these all confirmed government AI hallucinations?
No. The collection can include confirmed findings, acknowledged mistakes, disputed accounts, reported concerns and relevant discussion. A correction or withdrawal should be read for what it actually says; it does not necessarily acknowledge AI use or accept every criticism of the original material.
Can a single incident appear in several sources?
Yes. An original report, a corrected version, an official response and later proceedings may concern the same underlying matter. They should be read together. Separate records must not automatically be counted as separate incidents.
How can I suggest an addition or correction?
Please contact my clerks at Doughty Street Chambers, identifying the document or action, the relevant passage and the public source. Corrections, later responses and material clarifying whether AI was involved are all relevant to the research.
The questions behind the record
Reliability, accountability and public decision-making
My research asks what kinds of inaccuracies are reported, where they arise, how they are identified and what happens when they are challenged. It also considers how public bodies describe the use of AI and the steps taken to correct, explain or withdraw disputed material.
These questions sit within my wider work on AI law, public law, equality and human rights. They provide a basis for legal articles, reports, lectures and teaching about the developing use of AI in public administration. The tracker itself does not establish a legal conclusion about any individual record.
The public table is a concise guide to the material. Enquiries about the wider research should go through my clerks.
Related legal writing
AI tools and asylum decision-making
Read my discussion of a legal opinion on the Home Office’s use of AI in asylum decisions, alongside other developments in AI law.
Separate resources · Choose your next tracker
Looking for AI hallucination cases in court?
For detailed UK case records and figures, open the UK tracker. For selected judicial comments from around the world, open the international collection. Both are separate from this government tracker.
UK AI hallucination cases
The dedicated UK tracker, with case records, outcomes, judicial quotations, sources and research charts.
Open the detailed UK tracker →
International judicial comments
Read selected passages from judgments and decisions around the world, brought together to study the developing judicial discussion.
Open the international collection →More AI law trackers
Judicial AI use
Explore reported uses of AI by judges and related guidance and discussion.
Explore judicial AI →
AI equality, bias and discrimination
Research on AI’s risks of unequal treatment and its potential to reduce barriers and advance equality.
Explore the equality tracker →
Will AI replace lawyers?
Public statements about the future of lawyers, legal services and legal work.
Explore the debate →
International deepfake cases
Proceedings and public sources involving alleged or established synthetic images, audio and video.
Explore the deepfake database →AI law · Public law · Equality
Matthew Lee’s AI law practice
I accept instructions on many AI law matters, including AI issues arising within an existing case or another area of practice. My work includes legal advice, drafting pleadings, applications and other legal documents, advocacy and representation in disputes and proceedings involving AI.
If an AI legal issue arises, please contact my clerks at Doughty Street Chambers to discuss instructing me. Media, speaking, lecturing and teaching enquiries should also go through my clerks.
Contact my clerksLegal reports by area of practice
The AI Law Reports
I also founded The AI Law Reports, bringing together AI-related legal developments through familiar areas of practice, alongside legal reports and articles.
Important information
Matthew Lee accepts no responsibility or liability for errors, omissions or loss arising from use of or reliance on these resources. Nothing here is legal advice.
AI tools assist with research, organisation and updates. Records, summaries, references and other information may contain errors, omissions or inaccuracies and may be incomplete or out of date. Readers must independently check the original sources, their full context and current status and must not rely on this tracker when deciding how to act or as a substitute for legal advice.
Inclusion does not establish that an event occurred, that AI was used, that an allegation is true or that anyone acted improperly. A mention in a judgment or another public source may be sufficient to make a matter relevant to this research. Making an enquiry does not itself mean that instructions have been accepted.
