Britain's immigration inspectorate may open an investigation into the Home Office's use of generative AI to process asylum claims, WIRED has learned. The Independent Chief Inspector of Borders and Immigration has asked the Open Rights Group to supply more detail on the tools as it sets its inspection priorities for next year, campaigner Sara Alsherif said. The Home Office says AI only assists trained caseworkers and never makes decisions alone.

The department uses the Asylum Case Summarisation and Asylum Policy Search tools, built on OpenAI's 2023 GPT-4 model and Microsoft Foundry, to summarize interview transcripts and country-of-origin information. Piloting began in late 2024 as the government worked through a backlog of roughly 100,000 cases. ORG's six-month investigation found the share of ACS summaries unfit for use was well above the 9 percent error rate published from the pilot, and that APS summaries get no quality control.

Concerns have mounted over accuracy in decisions as consequential as returning someone to a country where they fear persecution. In July a UK judge accused the Home Office of citing an "AI hallucinated" document to refuse a Moroccan woman and her child's claim. Labour MP Bell Ribeiro-Addy said machines should not be making such calls, while ORG and the Public Law Project warn of cultural bias in GPT models and note the tools are absent from the government's public AI repository. The Information Commission is already investigating the department's refusal to disclose how staff use the tools.

Source: Wired