UK watchdog may investigate Home Office AI tools used in asylum cases

In short: A UK government watchdog is considering investigating how the Home Office uses OpenAI-based tools to help process asylum applications.

What happened

The Independent Chief Inspector of Borders and Immigration (ICIBI), which checks how the Home Office runs borders and immigration, has asked the Open Rights Group (ORG) for more information about the Home Office’s AI tools. ORG is a digital rights nonprofit that has spent six months trying to find out how the tools work.

According to documents ORG obtained, the Home Office has been piloting two tools since late 2024. One tool, Asylum Case Summarisation (ACS), creates summaries of interview transcripts. The other, Asylum Policy Search (APS), helps staff search and summarize country information that is used in decisions.

The Home Office says trained caseworkers make asylum decisions, and the AI tools are only used to help staff navigate records and policy. It also says it monitors performance and plans to publish updated evaluations.

Why it matters

These tools can save time, the government’s pilot estimated about an hour per case when used together. But critics say mistakes could be serious when decisions affect whether someone is sent back to a place they fear.

ORG says the share of AI summaries that were not fit for use was higher than the government’s published 9 percent error rate, often due to factual errors or misrepresentations. ORG also says there is no quality control for APS outputs. A judge also recently raised concerns after the Home Office allegedly cited an “AI hallucinated” document (meaning the system produced something that sounds real but is not).

There are also transparency concerns. The tools are not listed on the UK’s public register of algorithmic tools, and asylum seekers are not told when AI is used in their case. That can make it harder to challenge incorrect information, like trying to correct a bad summary of your own story.

Source: Wired