Liability for AI hallucinations

Liability for AI hallucinations designates the legal question of attributing the damage resulting from false, fabricated or misleading information, produced with an appearance of reliability by a generative artificial intelligence system. A hallucination can cause concrete harm: erroneous advice, defamation of a person, false financial or medical information, fabricated legal citation. The question then arises of who answers for the damage: the model publisher, the professional who deployed it without sufficient verification, or the user. This liability sits at the crossroads of product liability, the professional liability of the one relying on the tool, and defamation or consumer law depending on the case. Early court decisions have begun ruling on damage caused by false information generated by AI. For insurance, this emerging risk crosses several lines, professional indemnity, product liability, and calls for analysis of usage chains and the verification mechanisms put in place by deployers to limit their exposure.

A professional relying without verification on a legal citation fabricated by a generative AI can engage their professional liability toward the harmed client.

hallucination liability, responsabilité hallucination, responsabilité IA générative