The Independent Chief Inspector of Borders and Immigration is examining the Home Office’s use of two generative AI tools powered by GPT-4o to process asylum claims. Deployed through Microsoft Azure, Asylum Policy Search helps caseworkers find country-specific policy, while Asylum Case Summarisation condenses interviews into summaries. Post-rollout checks found a 22% failure rate across 203 summaries, compared with 9% during the pilot, and only 0.3% of the roughly 7,000 summaries generated each month were reviewed by subject matter experts. Applicants are neither notified that the tools are used nor allowed to view or challenge the summaries about them, which an Open Rights Group legal opinion says could render their use unlawful. The group is calling for their immediate suspension and independent audits, while the Home Office says trained caseworkers remain responsible for decisions.
Source: Read the original article
