ORG Investigation into use of ChatGPT in asylum decisions highlights errors, inaccuracies and Home Office incompetence
Open Rights Group (ORG) has carried out a six month investigation into two generative AI tools, being used by the Home Office in every asylum decision.
ORG, with the help of MPs Bell Ribeiro-Addy and Carla Denyer, has been trying to uncover how these tools have been developed, checked and used, as well as whether the government’s efficiency claims stand up.
The Home Office has refused to provide key information, including about staff training, the prompts being used, and how the tools are being tested for quality assurance. Their replies have been at times self-contradictory.
Our findings suggest that these tools have been rolled out without proper testing, quality assurance and training.
This careless approach could have catastrophic consequences for those people whose future is in the hands of Home Office decision-making.
The Use of ChatGPT in Asylum Decisions
Errors, Inaccuracies and Home Office Incompetence
Read the investigationBackground
The Home Office uses two AI tools in asylum casework, both built on a large language model (GPT-4o via Azure AI Foundry):
- The Asylum Case Summarisation tool (ACS) summarises asylum interview transcripts for decision-makers.
- The Asylum Policy Search tool (APS) is a chat-style interface for searching country policy information.
Both generate new text rather than retrieving existing text; they can filter, reframe or omit facts relevant to a decision.
APS was fully rolled out to all asylum decision-makers from July 2025. ACS was rolled out to all asylum caseworkers from April 2026.
“Deciding whether someone can or cannot seek refuge in the UK is one of the most serious and life-changing decisions that the state can make about an individual. The processes that lead to that decision must be transparent, fair and accountable.
But the Home Office continues to test AI on people seeking asylum in the UK without even telling them.
We urge the Government to regulate AI and outright ban its use in asylum applications, where the risk to rights are so high.”
Sara Alsherif
ORG Programme Manager – Migrant Digital Justice
Our findings
- Asylum claimants continue not to be told when AI is used to summarise their case and country information – something that is likely to be unlawful.
- Error rates are high, according to both the evaluation of the pilot and subsequent sample testing. These found that 9% and 15% of summaries respectively were not fit for purpose.
- There is no quality assurance for generated AI summaries from ACS.
- There is no quality check control for the generated text from the APS tool.
- Just 0.3% of summaries were checked by Home Office Subject Matter Experts.
- There are no plans for an independent audit of the system.
- The tools have not been added to the Algorithmic Transparency Recording Standard Hub.
- There has been no effort to bring the use of the tools into line with the AI Playbook for the UK Government.
