UK Uses AI to Summarise Asylum Interviews: One in Five Checked Summaries Failed

After the UK Home Office rolled out its AI summarization tool to all caseworkers in April 2026, 45 of 203 initial expert checks failed, a failure rate of about 22%, more than double the 9% recorded during the government's pilot. The digital rights group Open Rights Group (ORG) obtained the figures and has been asked for more information by the Independent Chief Inspector of Borders and Immigration, which is considering its 2027 inspection topics. The Home Office uses two AI tools, Asylum Case Summarisation (ACS) and Asylum Policy Search (APS), both running on OpenAI's GPT-4 family via Microsoft's Foundry platform. The Home Office stresses that trained caseworkers make the decisions and that the tools cannot be used in isolation. ORG also notes that asylum applicants are not informed when AI summarizes their interviews, and a commissioned legal opinion argues this may be unlawful. Meanwhile, a tribunal judge flagged possible AI use in an asylum refusal letter, and the data regulator is investigating the Home Office's refusal to disclose how the tools are used and staff trained.

(Source：International Business Times, Singapore Edition)