TDM and AI Training in the European Union – From ‘LAION’ to Possible Ways Ahead?

Abstract The first judgments have been rendered in the European Union that directly or indirectly deal with AI training, and the corresponding applicability, conditions and scope of the European exception to copyright for commercial text and data mining (TDM). These judgments serve as important impulses for the future interpretation of the provision and the current legal and policy debate on AI training and TDM in the EU. The most pressing question is how to define the requirements for a machine-readable opt-out pursuant to Art. 4(3) DSM Directive, even more so since Art. 53(1)(c) AI Act obliges providers of general-purpose AI models to put in place a policy to identify and comply with corresponding reservations of rights during the training process – even when carried out in a third country outside the EU. This is not only relevant from a doctrinal point of view. Rather, the interpretation and specification of and the interplay between Art. 4(3) DSM Directive and Art. 53(1)(c) AI Act are of key importance for future global AI training practice. At the same time, these provisions constitute the main anchoring point in EU copyright law for reflecting on the possible development of licensing markets for AI training material and/or possible de lege ferenda options under a legal policy perspective.