The UK’s Independent Chief Inspector of Borders and Immigration is considering whether to investigate the Home Office’s use of AI tools in asylum cases. The watchdog has asked the digital rights group Open Rights Group (ORG) for information as it plans next year’s inspection priorities, ORG campaigner Sara Alsherif told WIRED. The inspector told WIRED it could not confirm which topics it is actively considering, but said it knows about concerns over the tools.
The Home Office uses two tools built on OpenAI’s GPT-4 and Microsoft Foundry: Asylum Case Summarisation (ACS), which summarizes asylum interview transcripts, and Asylum Policy Search (APS), which finds and summarizes information about applicants’ home countries. The department began piloting them in late 2024, as it worked through a backlog of about 100,000 cases. Government estimates from the pilot said using the tools together could save about an hour per case. The Home Office says trained caseworkers make asylum decisions and the tools cannot decide claims on their own.
What are the concerns?
ORG’s six-month investigation found that the share of ACS summaries considered unfit for use was significantly higher than the 9 percent error rate reported in the pilot evaluation. It also found no quality control for APS summaries. The Home Office says it monitors the tools and has committed to publishing updated evaluations, but WIRED reports that the department has not shared details sought by ORG about how staff use the tools or are trained to use them.
The risks matter because caseworkers use the summaries and policy searches when assessing whether applicants face harm if returned to their country. In July, a UK judge criticized the Home Office for citing an “AI hallucinated” document when refusing an asylum claim by a Moroccan woman and her child. Researchers and lawmakers have also warned that errors or lost cultural context could affect decisions.
The tools are not listed publicly
The UK’s AI Playbook says government departments should disclose algorithmic tools used in decision-making and list them in a public repository. ACS and APS are not listed, and asylum applicants are not told when the tools are used in their cases, WIRED reports. ORG and the Public Law Project have raised concerns with the data regulator, which is investigating the Home Office’s refusal to provide information to ORG. Lawyers commissioned by ORG have said that preventing applicants from challenging incorrect summaries may breach data rights law.
The inspector has not confirmed that it will open a probe. A separate concern is that ACS and APS are among 11 high-risk government AI systems the Public Law Project says are missing from the public list. OpenAI told WIRED that it is working to improve model accuracy and plans to retire GPT-4 in January.
Why it matters
For people seeking asylum in the UK, AI-generated summaries and policy searches may inform the material caseworkers review, even though the Home Office says humans make the decisions. The watchdog is weighing a review, but it has not confirmed an investigation, and applicants are not told when the tools are used in their cases.
