UK Home Office faces investigation over AI tools used in asylum processing amid accuracy concerns

UK Home Office Under Scrutiny for AI in Asylum Cases

The UK Home Office may face investigation by the Independent Chief Inspector of Borders and Immigration (ICIBI) over its use of generative AI tools in asylum processing1

1

. The watchdog has requested additional information from digital rights nonprofit Open Rights Group (ORG) as it develops inspection priorities for next year1

1

. This follows a six-month investigation by ORG that involved 11 parliamentary questions, two ministerial letters, and 10 public records requests attempting to extract details from the Home Office about its AI tools in asylum processing1

1

.

Digital Hostile Environment Through AI Deployment

According to ORG's report titled "Automated Discrimination: Unpacking the Digital Hostile Environment," AI has been embedded across the UK Home Office in near-total secrecy2

2

. The technology is being used for tasks ranging from detecting potential sham marriages to determining asylum seekers' ages and checking phone contents of new arrivals2

2

. Officials have routinely invoked national security to stonewall freedom of information requests and redact disclosures2

2

.

Asylum Case Summarisation and Policy Search Tools

The Home Office uses two primary AI tools in asylum processing: Asylum Case Summarisation (ACS) and Asylum Policy Search (APS), both built on OpenAI's GPT-4 large language model and Microsoft Foundry1

1

. These tools summarize transcripts of interviews with asylum applicants and information about their countries of origin1

1

. The Home Office began piloting these generative AI tools in late 2024 as the Labour government attempted to clear a backlog of around 100,000 asylum cases1

1

. Government estimates indicate that when used together, the tools could save approximately an hour per case3

3

.

Accuracy Concerns and Error Rates Higher Than Reported

ORG found that the percentage of summaries generated by ACS deemed not fit for use—often because they contain factual errors or misrepresentations—is significantly higher than the published 9 percent error rate calculated in the pilot evaluation1

1

. More troubling, there is no quality control for the summaries produced by APS1

1

. In July, a UK judge accused the Home Office of citing an "AI hallucinated" document to refuse a Moroccan woman and her child's asylum claim1

1

. An inaccurate interview summary or country assessment could influence caseworkers even when the formal decision remains with a person3

3

.

Bias and Discrimination Risks in AI Tools

Labour backbench MP Bell Ribeiro-Addy, Sara Alsherif from ORG, and the Public Law Project worry that well-documented cultural bias in GPT models could lead to language barriers and the loss of cultural nuances in the summaries, potentially resulting in discriminatory decisions1

1

. "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," Ribeiro-Addy stated1

1

. The report emphasizes that AI is not neutral or accurate and can replicate bias and discrimination within society, particularly when a key driver for its use is anti-immigration policies and rhetoric2

2

.

Transparency Gaps and Missing Public Disclosure

Neither ACS nor APS appears on the public repository mandated by the UK's AI Playbook, which requires departments to disclose algorithmic tools used in decision-making processes3

3

. Public Law Project researchers identified ACS and APS among 11 high-risk government AI systems missing from the public repository3

3

. Applicants are not told when the tools are used in their cases3

3

. An independent legal opinion from Cloisters lawyers, commissioned by ORG, concluded that an inability to challenge an incorrect AI summary of personal information may breach UK data rights law3

3

.

Home Office Response and Ongoing Monitoring

A Home Office spokesperson emphasized that "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"1

1

. The department stated it regularly monitors the performance of these tools alongside established quality-assurance processes and has committed to publishing updated evaluations on their rollout1

1

. However, critics argue that scrutiny over AI use remains insufficient given the high-stakes nature of asylum decisions.

Calls for Regulation and Accountability

Open Rights Group is calling on the government to regulate AI and ban uses that are undemocratic and risk human rights violations, such as the use of OpenAI tools in asylum applications2

2

. The organization wants more transparency and stronger freedom of information laws, a narrower definition of national security, and a robust public register of government AI tools2

2

. Watch for whether the ICIBI formally launches an investigation and whether the Home Office will increase transparency around its use of AI tools in asylum processing. The broader question remains whether existing legal protections provide effective safeguards without clearer, enforceable regulation governing AI in high-stakes government decisions3

3

.