234 incidents on record · 2026HeadlightsIncident reports by Ellie Harris · Melbourne
5 new this weekLibrary last updated 14 October 2026
← The incident library
HD-INC-219
Immigration · United Kingdom · 2026 · General AI tools such as Copilot and ChatGPT were used inappropriately in asylum decision letters and interviews, although Copilot was not permitted for those processes, and the inspection that found it recorded no count of cases and no AI-specific recommendation

UK Inspectors Found AI Used Inappropriately in Asylum Decision Letters and Interviews

By Ellie Harris · Filed Inspection period June to December 2025

Alleged: Home Office (UK Visas and Immigration, Asylum Casework) 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.

UK Inspectors Found AI Used Inappropriately in Asylum Decision Letters and Interviews

What happened

The Independent Chief Inspector of Borders and Immigration examined the Home Office’s asylum casework from June to December 2025. The report was sent to the Home Secretary on 5 March 2026 and published on 25 June 2026.

The report says AI use in asylum casework “was in its infancy”. According to the report, approved tools, including the Asylum Policy Search tool, were available across all decision-making units by July 2025, although their effectiveness and reliability “were varied”.

The report says managers linked low awareness of the approved tools to “a clash with the rollout of Copilot, an AI tool that is not permitted for use in asylum interviewing or decision-making processes.”

What an auditable version would have shown

The report states: “Despite this, inspectors found there was evidence that AI applications like Copilot and ChatGPT had been used inappropriately in asylum decision letters and interviews.”

The report does not give the number of letters or interviews affected. It does not identify which tools were used in which cases or what they were used for. It records that “DMU managers told inspectors that these instances had been addressed”, but does not say how.

An auditable record would make it possible to show which decision letters involved an AI tool, which tool, and who reviewed the result.

Where the gap was

According to the report, 54 percent of asylum decisions received the top two quality scores in the 12 months to October 2025. This is a measure of overall decision quality, and the report does not link it to AI use.

The Chief Inspector made three recommendations. They covered organisational culture, putting quality alongside productivity, and improving feedback and lessons learnt. None of them is about AI.

A ConstraintGate is designed to check an action against declared rules, such as which tools may be used in a decision process, before it goes ahead.

A ConductRecord keeps which tool was used on each case, by whom and what was produced.

ConstraintGate and ConductRecord are Headlights designs.

What governance should have looked like

The Home Office’s published response accepted all three recommendations, with implementation due by December 2026. The response does not mention AI.

Where an organisation bans general AI tools from decisions about people, best practice would be to check for their use, record each case found, review the affected decisions and report what was done.

Failure Pattern: according to the Chief Inspector, AI applications such as Copilot and ChatGPT were used inappropriately in asylum decision letters and interviews, although Copilot was not permitted for those processes, and managers said the instances had been addressed.

Governance Principle: an organisation that bans a tool from a decision process should be able to show how it detects use of that tool and what it did about each case found.

The reference implementation of ConstraintGate 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.

Sent through Substack. 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 Home Office (UK Visas and Immigration, Asylum Casework) 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 October 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 →