UK immigration watchdog seeks information on Home Office asylum AI tools

Britain’s immigration watchdog has asked Open Rights Group for information about AI tools used in asylum casework, as it prepares next year’s inspection priorities. The request, reported by WIRED on October 27, brings the tools’ accuracy and safeguards into focus—but the watchdog has not confirmed an investigation or inspection topic.

Sara Alsherif, who leads Open Rights Group’s migrant digital justice program, told WIRED about the request. The Independent Chief Inspector of Borders and Immigration said it could not identify topics under active consideration before publishing its annual priorities. It acknowledged concerns about AI and pointed to its finding last year that one tool’s effectiveness and reliability varied.

Two shortcuts through an asylum file

The Home Office began piloting the systems in late 2024. According to documents obtained by Open Rights Group, they use OpenAI’s GPT-4 language model and Microsoft Foundry. Rather than issuing asylum decisions, the tools help caseworkers work through two different kinds of material: applicants’ own accounts and government information about their countries.

• Asylum Case Summarisation condenses interview transcripts. Those interviews record applicants explaining the persecution they experienced and why they fear returning home. Caseworkers use the summaries to speed up their review of that testimony.
• Asylum Policy Search helps officials find and summarize country-policy information. That material informs their assessment of whether an applicant would face a genuine risk of harm in their home country.

The attraction is speed. Government estimates from the pilot put the combined saving at around an hour per case when both tools were used. The Home Office says trained caseworkers still make asylum decisions, and that the AI tools cannot be used in isolation or decide claims themselves.

We regularly monitor the performance of these tools alongside our established quality-assurance processes, and have already committed to publishing updated evaluations on their roll out.

Home Office spokesperson, in an email to WIRED

Accuracy checks are part of the dispute

Open Rights Group’s findings followed a six-month investigation involving 11 parliamentary questions, two ministerial letters and 10 public-records requests. The group sought details about the tools that the Home Office had withheld, making access to information part of the dispute alongside the systems’ performance.

The group found that the share of interview summaries judged unfit for use was significantly higher than the pilot evaluation’s published 9 percent error rate. Such summaries often contained factual errors or misrepresentations. It also found no quality control for summaries produced by the policy-search tool—a separate concern from the interview-summary failure rate.

Applicants cannot see when AI enters their case

Asylum seekers are not told when these tools are used in their cases. The Home Office’s privacy notice does state that parts of its decision-making process may involve automation. That general notice sits alongside a more specific objection: applicants’ ability to challenge an incorrect AI account of their personal information.

An independent legal opinion commissioned by Open Rights Group, written by Cloisters lawyers, argues that being unable to contest an incorrect AI summary may breach UK data-rights laws. That is a legal assessment commissioned by the campaign group, not a finding that the Home Office has broken the law.

A separate disclosure investigation is already underway. The Information Commission is investigating the Home Office’s refusal to tell Open Rights Group how staff use the two tools and how they were trained to use them. Unlike the immigration inspector’s planning-stage request, that inquiry concerns an existing refusal to release information.