A UK government watchdog is weighing a possible investigation into the Home Office’s deployment of generative AI tools for processing asylum claims, WIRED reports. The move follows public scrutiny from parliamentarians and researchers who have questioned the accuracy of the tools and the department’s lack of transparency.
The Independent Chief Inspector of Borders and Immigration (ICIBI), which oversees the Home Office’s border, immigration and customs functions, has asked the digital‑rights charity Open Rights Group (ORG) for more details on the AI systems as it prepares its inspection agenda for next year, campaigner Sara Alsherif told WIRED. ORG’s conclusions stem from a six‑month inquiry that included eleven parliamentary questions, two letters to ministers and ten public records requests aimed at extracting information the Home Office has so far kept confidential.
ICIBI told WIRED it cannot confirm which topics are under active review before its annual inspection priorities are published, but it acknowledged awareness of concerns about AI use and referred to last year’s findings that the tools’ “effectiveness and reliability were varied.”
A Home Office spokesperson said in an email to WIRED, “Asylum decisions are made by trained caseworkers, with AI tools used only to assist in navigating interview records and policy information. They cannot be used in isolation and do not make decisions on claims. We regularly monitor the performance of these tools alongside our established quality‑assurance processes, and have already committed to publishing updated evaluations on their rollout.”
According to documents obtained by ORG, the department relies on two tools – Asylum Case Summarisation (ACS) and Asylum Policy Search (APS) – built on OpenAI’s 2023 GPT‑4 large language model and Microsoft Foundry. ACS summarises transcripts of interviews with asylum seekers, while APS helps locate and summarise information about the applicant’s country of origin. The Home Office began piloting the tools in late 2024 as the Labour government sought to clear a backlog of roughly 100,000 asylum cases and keep pace with the tens of thousands of new applications received each year. The department has reported that it halved the number of people waiting for an initial decision by the end of June, although the volume of appeals has risen sharply.
Researchers are uneasy about the tools’ accuracy, especially when they are used to decide whether a person should be returned to a country where they fear persecution. ORG found that the proportion of ACS summaries deemed unfit for use – often because they contain factual errors or misrepresentations – was considerably higher than the 9 percent error rate reported in the pilot evaluation. The group also noted that APS summaries lack any quality‑control mechanism.
The Home Office estimates that, when used together, the tools can save about an hour per case, allowing caseworkers to review interview transcripts and country policy information more quickly. In July, a UK judge accused the Home Office of relying on an “AI‑hallucinated” document to refuse asylum for a Moroccan woman and her child, arguing that the document incorrectly suggested her home country was safe.
Labour backbench MP Bell Ribeiro‑Addy told WIRED, “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.” The Home Office reiterated that AI is merely one tool among many and does not replace human decision‑makers.
Ribeiro‑Addy, Alsherif and the Public Law Project – which is also investigating the tools – warned that documented cultural bias in GPT models could create language barriers and erase cultural nuances in the summaries, potentially leading to discriminatory outcomes. An OpenAI spokesperson told WIRED that the company is continuously improving its models and will retire GPT‑4 in January.
Under the UK’s “AI Playbook,” government departments are required to be transparent about algorithmic tools used in decision‑making and to list them in a public repository. The Home Office’s case‑summarisation and policy‑search tools are not on that list, and asylum seekers are not informed when the AI systems are employed in their cases. “Migrant organisations, legal practitioners, not many of them know about the actual use of this tool,” Alsherif said. “So we need to let them know about this, let them know how it works, so they can challenge it.”
A legal opinion commissioned by ORG argued that the inability to contest an inaccurate AI summary of personal information may breach UK data‑rights laws. The Home Office’s privacy notice does acknowledge that parts of the decision‑making process may involve automation.
Both ORG and the Public Law Project have forwarded their findings to the Information Commissioner’s Office, which is currently probing the Home Office’s refusal to disclose how staff use and were trained on the two AI tools.