ACS & APS — the UK Home Office's AI Tools for Asylum Casework, With a Documented 9% Failure Rate

Facing an asylum backlog of more than 62,000 applications awaiting an initial decision as of September 2025, the UK Home Office piloted two generative-AI tools between May and December 2024: Asylum Case Summarisation (ACS), built on ChatGPT-4 to condense asylum interview transcripts, and Asylum Policy Search (APS), which retrieves and summarises country-of-origin policy guidance for caseworkers.
A Home Office research note reported that ACS saved decision-makers an average of 23 minutes per case and APS saved 37 minutes searching policy material — up to an hour combined — with neither tool appearing to change the quality of the final decisions in these small-scale pilots.
The same evaluation reported that 9% of ACS-generated summaries were flawed enough to be removed from the pilot, and that 23% of caseworkers lacked full confidence in the tools' outputs; civil-society reviewers separately noted the Home Office declined to disclose the prompts behind the tools or whether errors correlated with applicants' nationality or case complexity, and that asylum seekers are not told when AI is used in their case.
Both tools were designed to require a human decision-maker to review any AI output before a decision is made, and the Home Office's own research note recommends larger-scale evaluation before full rollout; ACS nonetheless proceeded toward a wider rollout in January 2026.

Read the full analysis: https://www.openrightsgroup.org/blog/saving-time-risking-lives-government-uses-ai-tools-to-inform-asylum-decisions/

Read the full analysis: https://www.ein.org.uk/news/home-office-expand-ai-use-asylum-decision-making-after-promising-pilot-results