Canadian man fined for submitting AI hallucinations as part of legal defense (www.theguardian.com)

A Quebec man, Jean Laprade, was fined C$5,000 after submitting court filings that relied on fabricated material generated by artificial intelligence — including eight non-existent citations and invented case law — while defending himself in a dispute over a miscontracted aircraft. The background: Laprade brokered a deal for helicopters and an airplane in Guinea that accidentally awarded him a far more valuable aircraft; an international arbitration ordered him to pay C$2.7M and the plane has been seized at Sherbrooke airport since 2019. Justice Luc Morin described Laprade’s use of AI as “highly reprehensible,” saying the fabricated excerpts were a serious breach that threatened the integrity of the tribunal, even as he acknowledged the defendant’s lack of counsel and the seductive utility of AI. The case is a concrete reminder of the technical limits and legal risks of large language models: hallucinations — confidently produced but false or nonexistent facts and citations — can mislead courts if not rigorously verified. The judge cited a 2023 notice urging the legal profession to subject AI-generated material to “rigorous human control,” and warned that AI’s “intoxicating promises” will test judicial vigilance. For the AI/ML community this underscores the practical need for provenance, citation verification, and tools to flag hallucinations when LLMs are used in high-stakes, adversarial contexts.