Legal Opinion: The Use of Artificial Intelligence Tools in Asylum Cases

Key Takeaways

• AI summaries inaccurate 9% of pilot cases
• Asylum seekers unaware of AI usage
• Human control mechanisms not documented
• No impact assessments published for AI tools
• Regulators yet to scrutinize Home Office AI

Summary

Open Rights Group released a legal opinion examining the UK Home Office’s use of generative AI tools—ACS and APS—in refugee status determinations. The opinion highlights that the tools produced inaccurate summaries in up to 9% of cases, lack transparent oversight, and appear to breach several UK AI Playbook principles and domestic legal duties. It warns that asylum seekers are not informed about AI involvement, limiting procedural fairness and risking violations of human rights, data protection, and equality duties. The report calls for robust AI assurance, impact assessments, and regulator scrutiny to align practice with international ethical standards.

Pulse Analysis

The UK government has embraced artificial intelligence to streamline the asylum adjudication process, positioning the Home Office’s ACS (Applicant Content Summariser) and APS (Country Profile Search) tools as efficiency boosters. These generative AI systems sit within the broader UK AI Playbook, which mirrors UNESCO, OECD and Council of Europe ethical principles such as fairness, transparency, and accountability. While the Playbook outlines procedural safeguards—human oversight, risk assurance, and impact assessments—public documentation shows a stark gap between policy and practice, especially in high‑stakes migration contexts where decisions affect fundamental rights.

Critical analysis of the legal opinion reveals concrete shortcomings. Pilot data indicate a 9% error rate for ACS summaries and notable confidence gaps for APS, yet the Home Office has not disclosed quantitative benchmarks or validation methods. Decision‑makers receive AI‑generated text without access to original source material, eroding meaningful human control and contestability. Moreover, asylum seekers remain uninformed about AI involvement, contravening procedural fairness under Article 3 of the Human Rights Act and breaching data‑protection obligations. The absence of published Data Protection Impact Assessments, Equality Impact Assessments, and algorithmic transparency registers further signals non‑compliance with both domestic law and the AI Playbook’s Principle 2 and Principle 7.

For civil society and regulators, the opinion underscores an urgent need for oversight. Independent bodies such as the Independent Chief Inspector of Borders and Immigration must audit AI deployments, enforce AI assurance frameworks, and demand full disclosure of training data, model limitations, and mitigation strategies. Aligning AI use with established ethical standards not only safeguards refugee rights but also preserves public confidence in the immigration system. As AI technologies evolve, robust governance will be essential to prevent systemic bias, ensure accountability, and uphold the rule of law in the UK’s asylum process.