Ayinde Case: "Supported Arguments with Non-Existent Precedents"
Al-Haroun Case: "Fictional Citations Fabricating Even Judge Names"
June 6, 2025: UK High Court released two unusual rulings that strongly warned the world how dangerous legal document creation using AI can be. The first case was a lawsuit a citizen filed against local government over housing support; the second was international litigation over financial contracts worth tens of millions of pounds. The court discovered shocking facts: documents submitted by lawyers in both cases contained numerous fake precedents that do not exist. Cases that appeared to be legal foundations were actually fictions plausibly fabricated by AI, with some citation numbers not even matching actual cases. The court warning: AI-generated legal citations require mandatory verification before submission; lawyers have a professional duty to verify the existence and accuracy of every cited case; submission of AI-generated citations without verification may constitute professional misconduct; courts will impose sanctions (costs orders, professional referrals) for submission of false citations. Global pattern: this UK warning follows similar incidents in the US (Mata v. Avianca 2023 -- a New York federal judge sanctioned lawyers who submitted ChatGPT-fabricated citations); Canada, Australia, and New Zealand courts have all issued similar warnings. The AI hallucination in legal context: large language models trained on legal text learn the form of legal citation without reliably verifying case existence; when asked to find supporting precedent, models confidently generate plausible-sounding cases that do not exist; the formal structure of the fabrication (complete with judge names, holdings, and quotations) makes them difficult to identify without actively searching for the cited case. The professional liability implications: law firm liability for AI-hallucinated citations is now established precedent; legal malpractice insurers are developing specific AI use policies; bar associations are developing guidance on AI-assisted legal work that preserves professional responsibility standards.
