Via jurist.org
Upper tribunal judges compared reliance on an unlocatable AI-cited document to 'bogus evidence,' raising fresh questions about AI accountability in high-stakes government decisions
Here’s a sentence that should terrify anyone building AI systems for institutional use: a UK court just compared AI-generated evidence used in an asylum case to “bogus evidence.” The Upper Tribunal couldn’t even find the document the Home Office cited, which is the legal equivalent of turning in a book report on a novel that doesn’t exist.
What actually happened #
UK Upper Tribunal judges reviewing an immigration case found that a document cited by the Home Office simply could not be located. Not misfiled. Not redacted. Not buried in bureaucratic limbo. It apparently didn’t exist.
The tribunal’s response was blunt, calling the reliance on this phantom document “analogous to relying on bogus evidence.”
This wasn’t an isolated incident either. Between 2025 and 2026, multiple rulings from the Upper Tribunal addressed fictitious case citations generated by AI systems, including ChatGPT and Google’s AI Mode, appearing in asylum and immigration filings prepared by legal representatives.
The tribunal also ruled that up Home Office decision letters to open-source AI tools like ChatGPT waives confidentiality and legal privilege.
By July 2026, the Crown Prosecution Service had issued apologies for citing AI-generated non-existent legal authorities in extradition cases.
The bigger AI accountability gap #
The EU’s AI Act, which began phased enforcement in 2025, attempts to classify AI applications by risk level, with immigration decisions falling into the “high-risk” category requiring human oversight and transparency. But as the UK cases demonstrate, having humans in the loop doesn’t help if those humans trust the AI output without independent verification.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our