The UK’s independent border inspector is considering whether to investigate two generative-AI tools used by the Home Office during asylum processing. The inspector asked the Open Rights Group for information as it prepares next year’s priorities, although it has not confirmed that an inquiry will proceed.
The Asylum Case Summarisation tool condenses applicant interviews, while Asylum Policy Search helps caseworkers navigate country-policy information. Both use OpenAI’s 2023 GPT-4 through Microsoft Foundry. The Home Office began pilots in late 2024 while working through a backlog of roughly 100,000 cases.
Officials say trained caseworkers make decisions and the tools cannot be used alone. A pilot estimated that using both could save about an hour per case. The department says it monitors performance and will publish updated evaluations.
Rights researchers argue that the available evidence is incomplete. Open Rights Group says the share of summaries judged unfit for use is substantially higher than the published 9 percent pilot error rate, while the policy-search tool lacks equivalent quality control. A judge also accused the department in July of citing an AI-invented document in an asylum refusal.
Because errors can affect whether someone is returned to a country where they fear persecution, time savings are not enough. A meaningful review would need current error rates, representative tests, records of human corrections and clear disclosure of how outputs influence each decision.