The Pulse
UK Watchdog Considers Probe Into OpenAI Asylum Tools
The UK’s borders inspector is weighing whether to inspect Home Office use of AI tools that summarise asylum interviews and search country information. Open Rights Group says later checks found more summaries unfit for use than the Home Offi
AI.info Team · Written with AI · fact-checked ·
“We’re dealing with people’s lives. We’re talking about people that have faced some of the worst circumstances, and we’re meant to be making decisions based on facts and on understanding harm. I just don’t think it’s appropriate that machines are doing that.”
Bell Ribeiro-Addy, Labour backbench MP
The UK’s Independent Chief Inspector of Borders and Immigration is considering whether to inspect the Home Office’s use of generative AI in asylum casework. The inspector has asked digital-rights group Open Rights Group (ORG) for more information while planning next year’s inspection priorities, WIRED reported on October 27. The watchdog has not confirmed that it will open an inspection.
The inspector is weighing a review, not announcing one
The Independent Chief Inspector of Borders and Immigration monitors the Home Office’s work on borders, immigration, asylum and customs. It told WIRED it could not confirm which subjects were under active consideration before publishing its annual inspection priorities, but acknowledged concerns about AI tools and pointed to an earlier finding that one tool’s “effectiveness and reliability were varied.”
ORG’s six-month investigation prompted the request for information. The group says it made 11 parliamentary inquiries, sent two letters to ministers and filed 10 public-records requests seeking details about how the systems were developed, checked and used. Separately, the Information Commission is investigating the Home Office’s refusal to provide ORG with information about staff use and training, WIRED reported.
Two tools summarise interviews and country information
The Home Office uses the Asylum Case Summarisation tool, or ACS, to condense interview transcripts, and Asylum Policy Search, or APS, to find and summarise information about conditions in applicants’ countries of origin. Decision-makers use those materials when assessing claims, including whether an applicant faces a risk of harm if returned.
The Home Office told WIRED that trained caseworkers make asylum decisions and that AI tools assist with navigating interview records and policy information; staff cannot use them in isolation. The department’s pilot evaluation likewise described both systems as aids, not substitutes for human decision-making. ORG says APS reached all asylum decision-makers in July 2025 and ACS all asylum caseworkers in April 2026.
ORG’s later checks found a higher share of faulty summaries
The Home Office’s 2025 evaluation found that 9 percent of ACS summaries in its pilot were inaccurate or missing information. Those summaries were removed from the pilot, and the affected cases proceeded through the usual process. The government’s pilot also estimated that ACS saved 23 minutes per case reviewing interview transcripts and APS saved 37 minutes per case searching for country-policy information; it cautioned that the trials were small-scale.
ORG says its later sample testing found that 15 percent of summaries were not fit for purpose. It also reports that just 0.3 percent of summaries generated between April and July 2026 were checked by Home Office subject-matter experts, and that ACS summaries and APS-generated text lack quality-control checks. The measures describe different stages and checks: the Home Office’s pilot result is not a direct evaluation of the systems after full rollout.
Applicants may not know when AI is used
ORG says asylum applicants are not told when the tools summarise their cases or country information, and that the systems do not appear on the government’s public register of algorithmic tools. The group argues that applicants need enough information to challenge errors in material about their personal circumstances. Lawyers commissioned by ORG have separately raised concerns about possible data-rights violations, but those concerns are not a finding by a court or regulator.
The Home Office says it monitors the tools and has committed to publishing updated evaluations as rollout continues. For now, the key oversight question is narrower than whether a probe is underway: the border inspector has requested evidence while setting next year’s priorities, but has not said whether the two tools will be included.