ABSTRACT
This article explores the potential liability of generative AI (GenAI) operators for reputational harm under the EU legal framework, with a focus on three possible legal grounds: defamation, product liability, and non-contractual liability. National defamation law faces significant challenges in addressing AI-generated content in the absence of human intent or editorial control. While the new Product Liability Directive classifies AI systems as products, it excludes damages resulting from violations of personality rights from its scope of application. Accordingly, Member States’ non-contractual liability rules may offer the most viable path for claimants, despite the challenges of proving fault and the risk of conflicting judgments between jurisdictions. Since GenAI operators cannot foresee specific defamatory outputs, they should be held liable only when they know about the illegality – usually after a notification – and fail to take reasonable steps to block the output. These steps could include implementing machineunlearning techniques or filters. This would mean subjecting GenAI operators to a liability regime similar to that applicable to online platforms for hosting defamatory content.
€
Laura Herrerías Castro, Liability of Generative AI for Defamation, European Review of Private Law volume 34, issue 2 (2026).
Leave a Reply