Table of Contents
The Independent Chief Inspector of Borders and Immigration (ICIBI) has asked the digital rights nonprofit Open Rights Group (ORG) to provide more information regarding AI tools used by the UK Home Office. As reported by WIRED, the inquiry comes as the oversight body prepares its inspection priorities for next year, following a six-month investigation by ORG that involved numerous parliamentary questions, ministerial letters, and public records requests.
AI Tools in Asylum Processing
The Home Office has deployed the Asylum Case Summarisation (ACS) and Asylum Policy Search (APS) tools, which are built on OpenAI’s GPT-4 large language model and Microsoft Foundry. These systems summarize interview transcripts of asylum applicants and compile country-of-origin information. The immigration authority started piloting these tools in late 2024 to help address a backlog of approximately 100,000 asylum cases while managing ongoing new applications.
Government estimates indicate that using the tools together can save caseworkers about an hour per case by speeding up transcript reviews and policy searches. However, a Home Office spokesperson told WIRED via email that asylum decisions are handled by trained caseworkers, with AI used solely to assist in navigating records and policies. The spokesperson added that AI tools cannot be used in isolation, do not make final decisions, and are monitored regularly alongside quality-assurance processes.
Accuracy and Accountability Concerns
Researchers have expressed serious concerns about the reliability of the AI systems in high-stakes asylum determinations. ORG found that the proportion of ACS summaries deemed unfit for use due to factual errors or misrepresentations is significantly higher than the 9 percent error rate stated in the pilot evaluation. Furthermore, the organization discovered that APS summaries lack any formal quality control.
In July, a UK judge accused the Home Office of citing an AI-hallucinated document to reject an asylum claim from a Moroccan woman and her child. Labour MP Bell Ribeiro-Addy told WIRED that using machines for such critical decisions involving human safety is inappropriate. Meanwhile, an OpenAI spokesperson highlighted ongoing efforts to improve model accuracy and noted that GPT-4 is scheduled for retirement in January.
Transparency and Regulatory Scrutiny
Under the UK’s AI Playbook, government departments must remain transparent about algorithmic tools and list them on a public repository. Neither the case summarization nor the policy search tools appear on that list, and asylum seekers are not informed when AI is used in their cases. Sara Alsherif, who leads ORG’s migrant digital justice program, told WIRED that the lack of awareness prevents migrants and legal practitioners from challenging the outputs, potentially breaching UK data rights laws according to an independent legal opinion.
Both ORG and the Public Law Project have shared their findings with the Information Commission, which is currently investigating the Home Office for refusing to disclose how staff use and are trained on the two AI tools. Caroline Selman, a senior research fellow at the Public Law Project, told WIRED that ACS and APS are just two of 11 high-risk AI systems her team identified across the government that are missing from the public repository.
The ICIBI informed WIRED that it cannot confirm specific topics under active consideration prior to publishing its annual inspection priorities, though it acknowledged awareness of concerns surrounding AI tool usage and noted past findings regarding varied reliability.
