The UK government’s use of generative artificial intelligence to help process asylum applications could face an investigation by the Independent Chief Inspector of Borders and Immigration (ICIBI), WIRED reported. The possibility follows public concerns from lawmakers and researchers about the accuracy of the tools and the Home Office’s lack of transparency about how they are being used.
Watchdog seeks information about AI tools
The ICIBI, which oversees the Home Office’s work on borders, immigration and customs, has asked the digital rights nonprofit Open Rights Group (ORG) to provide more information about the tools as it considers topics for next year’s inspections, campaigner Sara Alsherif told WIRED.
ORG’s findings are based on a six-month investigation that included 11 parliamentary questions, two ministerial letters and 10 public records requests seeking information from the Home Office. The department withheld details about the tools.
The ICIBI told WIRED that it could not confirm which topics were actively being considered before its annual inspection priorities are announced. However, it said it was aware of concerns about the use of AI tools.
Last year, concerns were raised that the tools had “variable validity and reliability.”
“Decisions on asylum applications are made by trained caseworkers and AI tools are only used to assist in navigating interview records and policy information. AI tools cannot be used in isolation and do not make decisions on applications,” a Home Office spokesperson told WIRED in an email.
“We regularly monitor the performance of these tools alongside our established quality assurance processes and are already committed to publishing updated assessments upon implementation,” the spokesperson added.
How the Home Office uses the tools
Documents obtained by ORG indicate that the Home Office is using two tools: Asylum Case Summary (ACS) and Asylum Policy Search (APS). Both were built using OpenAI’s 2023 GPT-4 large language model and Microsoft Foundry.
ACS summarizes asylum interview records, in which applicants describe how they were persecuted in their home countries and why they fear returning. APS acts as a search assistant, finding and summarizing country-of-origin information that caseworkers use to assess whether an applicant faces a real risk of harm.
Immigration and law enforcement officials began piloting the tools in late 2024, as the Labour government sought to address the asylum backlog. The UK handles around 100,000 asylum applications and receives tens of thousands of new applications each year.
The number of people waiting for an initial decision was cut in half in the year to June, according to the source article. However, the number of appeals against decisions has also increased sharply.
Researchers question accuracy and oversight
Researchers have raised concerns about the reliability of AI systems used in high-stakes decisions, including decisions about whether someone should be returned to a country where they fear persecution.
ORG found that the proportion of ACS summaries considered unsuitable for use was significantly higher than the 9 percent error rate published in a pilot evaluation. Summaries were deemed unsuitable in many cases because they contained factual errors or misstatements.
ORG also found that APS-generated summaries did not have a quality-control process.
Government estimates from trials indicate that using the two tools together can save about one hour per case. The tools are intended to help asylum decision-makers work more quickly, but decisions are formally made by trained caseworkers, according to the Home Office.
Judge cites disputed AI-generated material
In July, a British judge accused the Home Office of citing an “AI illusion” document when rejecting the asylum application of a Moroccan woman and her child. The decision said it would be safe for them to return to Morocco.
Source: www.wired.com