Generative AI – challenges for trade mark law

Abstract

This exploratory paper examines the regulatory challenges that generative artificial intelligence (GenAI) poses to European trade mark law. GenAI systems, trained on vast datasets collected via web scraping, frequently incorporate registered trade marks without the consent of their owners. The article analyses whether such use during model training constitutes relevant “use in the course of trade” for the purposes of trade mark infringement under the Trade Mark Directive and the EU Trade Mark Regulation, arguing that existing legal frameworks – designed for human, intentional and deterministic uses of distinctive signs – may be ill-suited to the probabilistic and opaque nature of GenAI. Particular attention is given to the heightened vulnerability of trade marks with reputation, which are deeply embedded in AI training datasets and whose dilution and parasitic exploitation by GenAI outputs raise novel evidentiary and conceptual difficulties. Drawing on recent litigation – notably Getty Images v. Stability AI (UK) and The New York Times v. Microsoft & OpenAI (US) – the paper concludes that the current legal framework struggles to accommodate these challenges and considers whether legislative reform or a purposive reinterpretation of existing provisions, potentially complemented by the EU AI Act’s transparency and training-data disclosure requirements, could provide more effective protection for trade mark owners.