AI tool used in thousands of asylum claims failing basic accuracy tests

By the Editorial Team · Friday, October 9, 2026

The UK Home Office has faced criticism after its AI tool for summarizing asylum interview transcripts was found to frequently produce inaccurate and inconsistent results, affecting decision-making quality. This raises concerns about the balance between speeding up claim processing and ensuring fair, transparent outcomes, especially as appeals rise and calls for greater oversight intensify.

The UK Home Office has come under scrutiny for its use of an artificial intelligence (AI) tool designed to accelerate the processing of asylum claims, amid concerns over the tool’s accuracy and transparency. The AI system, known as the Asylum Case Summarisation (ACS) tool, has been used since April 2026 to summarise lengthy interview transcripts from asylum seekers, aiming to reduce delays in decision-making.

Internal checks conducted by Home Office experts revealed that the tool frequently produced inaccurate summaries. Of 259 reviews carried out between April and July 2026, 22 percent of primary reviews and 54 percent of secondary reviews failed to meet accuracy standards. Failures included multiple factual errors, significant inconsistencies with interview content, and summaries deemed unhelpful for caseworkers. Asylum seekers are not informed when AI is used in processing their claims.

The ACS tool employs OpenAI’s GPT-4 model and was initially piloted alongside another AI tool, the Asylum Policy Search (APS), which assists decision-makers by summarising country-of-origin policy information. While the ACS system reportedly saved decision-makers an average of 23 minutes per case, Home Office officials emphasized that AI does not replace human judgment. Decision-makers are required to review all documents pertinent to each claim and have discretion over whether to use AI-generated summaries.

The use of AI in this sensitive context has drawn criticism from MPs, charities, and rights groups, who warn that prioritising speed over accuracy risks miscarriages of justice and undermines fundamental legal standards. Labour MP Bell Ribeiro-Addy has expressed concern over what she described as limited human oversight and widespread inaccuracies in the tool’s output. Ben Maguire, the Liberal Democrat shadow attorney general, cautioned against AI substituting genuine human scrutiny.

Additional voices highlighted the challenges unique to asylum cases, including the trauma and language barriers faced by applicants during interviews. Representatives from advocacy organizations stressed that errors in interview records fed into AI could lead to unreliable outcomes and heightened distrust in the asylum system. They argued that robust human review must remain central to decision-making, especially given the potentially life-altering consequences.

The Home Office stated that all asylum decisions are made by trained human decision-makers who use AI tools solely as aids for navigating complex information. It maintained ongoing quality assurance and monitoring of the AI systems to safeguard the integrity of the process.

As of June 2026, the initial backlog of asylum claims has reportedly decreased, but appeals have increased, with over 90,000 cases pending. The government faces pressure to balance efficiency with fairness while contending with the backlog and legal obligations. Calls for greater transparency continue, with some critics noting that detailed information about AI’s role has only emerged after multiple parliamentary inquiries and freedom of information requests.

This article was reported by synthesizing coverage from multiple independent news outlets to present a balanced, unbiased account. Read about our editorial standards. Spotted an error? Let us know.