Home Office asylum AI flunks one in five accuracy checks
The Home Office is using artificial intelligence tools to process asylum cases more quickly, but a new report from the Open Rights Group has revealed significant problems with accuracy and effectiveness. According to the investigation, more than one in five AI-generated interview summaries failed accuracy checks after the system was rolled out. Despite ministers' promises of faster asylum decisions, the technology has sparked serious concerns about whether it is working as intended and whether adequate safeguards are in place.
An expert review found that 45 out of 203 summaries failed initial checks, a failure rate of 22 per cent compared with just 9 per cent during the pilot phase. The situation is complicated by the fact that only 0.3 per cent of all summaries have been reviewed, and real-world data is troubling: since April 2026, the average time from interview to decision has increased by 29 per cent whilst the number of decisions made has fallen by 42 per cent. The Open Rights Group warns that asylum seekers are not informed when AI is used in their cases and cannot inspect or challenge the summaries, whilst the removal of experienced human reviewers from the process has weakened rather than strengthened oversight.
- More than one in five asylum AI interview summaries fail accuracy checks.
- Processing times have worsened since the AI tools were introduced.
- Asylum seekers cannot see or challenge AI-generated summaries about their cases.
New here? Start with this
The Home Office is using artificial intelligence to help process asylum cases more quickly. The system automatically generates written summaries of asylum interviews, intended to speed up decision-making on applications from people seeking refuge in the UK.
A report from the Open Rights Group has found significant problems with how well the AI is working. The AI-generated summaries have failed accuracy checks at much higher rates than when the system was first tested, and in practice, asylum decisions have slowed rather than sped up since the technology was introduced.
The investigation has raised concerns about whether asylum seekers have adequate protections when AI is used in their cases. People applying for asylum are not told when AI is involved and cannot challenge the system's summaries, whilst experienced human reviewers have been removed from the process, weakening oversight of decisions that affect vulnerable people's lives.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Proponents argue that the asylum backlog has created an untenable situation where cases languish for years, causing genuine harm to vulnerable people waiting in limbo. AI-assisted summaries could accelerate routine administrative tasks, freeing experienced human reviewers to focus on complex decisions. The pilot phase showed 91 per cent accuracy, and they contend that scaling challenges are surmountable through refinement. With resource constraints, deploying technology to handle routine processing allows limited staff to concentrate on substantive matters, potentially improving overall throughput.
The case against
Critics contend that asylum cases involve fundamental questions of safety and human rights that demand rigorous accuracy and personal scrutiny. A 22 per cent failure rate represents an unacceptable error margin when individuals' protection depends on correct processing, and the real-world data—longer processing times and fewer decisions since deployment—suggests the system has made matters worse. They argue that asylum seekers, as a vulnerable population, deserve full transparency about how their cases are handled and the right to challenge inaccurate summaries. Without meaningful human oversight and accountability mechanisms, the pursuit of speed undermines the fairness the system owes to people seeking refuge.