The UK government’s Home Office may come under official investigation for its use of AI tools powered by OpenAI’s GPT-4 technology to support asylum claim reviews, amid mounting worries about errors and opaque deployment.
- Home Office uses GPT-4-based tools to assist asylum case reviews since late 2024
- Investigations triggered by error rates and lack of transparency in AI summaries
- Concerns over cultural bias and AI ‘hallucinations’ affecting refugee decisions
What happened
The UK Home Office employs AI tools built on OpenAI’s GPT-4 model to summarize interview transcripts and policy information related to asylum cases. These AI systems—Asylum Case Summarisation (ACS) and Asylum Policy Search (APS)—were introduced in a pilot phase starting in late 2024 to help clear a backlog of around 100,000 asylum applications and manage new claims more efficiently. The tools aim to reduce casework time by about an hour per case, helping decision-makers navigate complex information faster.
However, the Independent Chief Inspector of Borders and Immigration has asked for more information following a six-month investigation by digital rights group Open Rights Group. They uncovered higher-than-reported error rates in the AI-generated summaries and a lack of quality control for policy search outputs. Additionally, a UK judge recently criticized the Home Office for relying on AI-generated information that was found to be inaccurate when denying an asylum claim. This has prompted calls for closer examination of the AI’s role and effectiveness in this sensitive area.
Why it matters
Asylum decisions profoundly impact the lives of individuals fleeing persecution, meaning accuracy and fairness are paramount. Critics highlight that errors, misrepresentations, or ‘hallucinated’ AI data could lead to wrongful denials, putting vulnerable people at risk of harm if returned to unsafe conditions. The use of AI also raises questions about transparency and the extent to which these tools influence decisions alongside human caseworkers.
Further concerns include the potential for cultural biases embedded in GPT models to distort or oversimplify asylum seekers’ testimonies and policy contexts. This can undermine the nuanced understanding required for complex immigration decisions. Lawmakers and advocacy groups worry that overreliance on AI without rigorous oversight and ongoing evaluation could compromise justice and the integrity of the UK’s asylum process.
What to watch next
The Independent Chief Inspector of Borders and Immigration is expected to determine whether to launch a formal investigation into the Home Office’s AI practices as part of its upcoming annual inspections. Stakeholders will closely monitor any forthcoming reports, looking for concrete findings on the tools’ reliability, impacts, and compliance with legal standards. The Home Office has pledged to publish updated evaluations of these AI systems and to maintain human decision-making primacy.
Observers should also watch parliamentary responses and potential regulatory actions related to AI use in public sector decision-making. The case highlights broader challenges of integrating generative AI within government services—balancing efficiency gains against ethical, legal, and human rights risks. Developments here could influence future policies on AI-assisted immigration processing and wider public sector AI governance.