UK Home Office Used AI-Hallucinated Document to Deny Asylum Claim, Judge Finds

The UK Home Office has been accused of relying on an AI-hallucinated document to refuse an asylum claim, according to a July 2026 ruling by the Upper Tribunal (Immigration and Asylum Chamber). The case involves a Moroccan woman and her child who fled forced underage marriage, rape, and threats from her convicted criminal husband. The Home Office rejected her asylum application, citing a country policy information note (CPIN) allegedly confirming Morocco would be safe for her return. However, no such CPIN could be found.

Key Findings from the Ruling

• Missing document: The Home Office's refusal letter referenced a "country information note Morocco July 2021" but the Upper Tribunal could not locate any such document. The Home Office's own country policy team confirmed it could not confirm the document's existence.
• Judge's conclusion: The judge stated the refusal letter "bears hallmarks consistent with the use of artificial intelligence" and the reference may be a result of an AI "hallucination." He called relying on a non-existent document "analogous to relying on bogus evidence."
• Shift from CIN to CPIN: After the Guardian contacted them, the Home Office produced an internal, unpublished country information note (CIN) from 2018. The refusal letter originally cited a CPIN—considered authoritative by courts—but later references shifted to the CIN. No CPIN on Morocco has been published since 2017.
• Content of the internal document: The CIN does not conclude Morocco is safe. It states Morocco granted 40,000 child marriages in 2018 (one-fifth of marriages that year), does not criminalize marital rape, and may criminalize women reporting rape under adultery laws.

Practical Implications for AI Use in Government

This case marks a concrete legal finding of AI hallucination causing real-world harm in a high-stakes administrative decision. Developers working on AI-assisted document generation or retrieval systems should note:

• AI-generated citations must be verified against authoritative sources before use in any legal or official context.
• Over-reliance on LLM outputs without human validation creates liability—here, it potentially contributed to a wrongful denial of asylum.
• The shift from citing an internal CIN to a CPIN suggests possible post-hoc rationalization after the hallucination was exposed.