A high-court judge has rebuked the Home Office after finding that an asylum refusal letter relied on information that appears never to have existed, raising fresh doubts about the department’s experimental use of large-language-model tools in case-work. In the judgment handed down on 28 July 2026, Mr Justice Millwood noted that the refusal quoted a ‘Morocco country information note – July 2021’ containing statistics on LGBTQ+ prosecutions. No such note was ever published, leading the judge to suggest the data had been generated by an ‘AI hallucination’—a fabricated citation produced by a generative-AI assistant. The Home Office has acknowledged it is piloting artificial-intelligence drafting aids to speed up decision letters under its ‘Beta-Decision’ programme launched last December. Officials insisted that human case-workers remain “fully responsible” for checking content, yet the ruling exposes gaps in quality-assurance processes. The asylum seeker—a Cameroonian national who arrived in the UK by air in 2024—has now had the refusal quashed and the case remitted for reconsideration. For global mobility stakeholders—including employers that sponsor humanitarian routes such as the Community Sponsorship Scheme—the episode carries broader implications. If unverified AI-generated material finds its way into visa or status decisions, the risk of unlawful refusals increases, potentially delaying workforce deployment and undermining confidence in the immigration system. Multinationals operating compliance programmes should be prepared for additional due-diligence steps when challenging adverse decisions. Legal experts predict heightened scrutiny from oversight bodies. The Independent Chief Inspector of Borders and Immigration is already conducting a thematic review of AI usage across UKVI, Border Force and HM Passport Office, with a report expected in autumn 2026. Parliament’s Home Affairs Committee chair has called for an immediate pause to any AI roll-out until robust safeguards, including source-citation verification and audit trails, are in place. Practically, mobility advisers should monitor whether standard decision templates change and whether refusal notices include verifiable hyperlinks to country policy and information notes. Employers supporting staff through protection-based routes or dependent partner applications may wish to budget for potential legal challenges and build extra lead-time into assignment schedules until the Home Office clarifies its AI governance framework.
Source: The Guardian
