Generative AI, Copyright and the AI Act (v.2)

Published 1 November 2024. This is a revised and extended version of a paper initially published in August 2024. This paper examines the copyright-relevant rules of the recently published Artificial Intelligence (AI) Act for the EU copyright acquis. The aim of the paper is to provide a critical overview of the relationship between the AI Act and EU copyright law, while highlighting potential gray areas and blind spots for legal interpretation and future policy-making. The paper proceeds as follows. After a short introduction, Section 2 outlines the basic copyright issues of generative AI and the relevant copyright acquis rules that interface with the AI Act. It mentions potential copyright issues with the input or training stage, the model, and outputs. The AI Act rules are mostly relevant for the training of AI models, and the Regulation primarily interfaces with the text and data mining (TDM) exceptions in Articles 3 and 4 of the Copyright in the Digital Single Market Directive (CDSMD). Section 3 then briefly explains the AI Act’s structure and core definitions as they pertain to copyright law. Section 4 is the heart of the paper. It covers in some detail the interface between the AI Act and EU copyright law, namely: the clarification that TDM is involved in training AI models (4.1); the outline of the key copyright obligations in the AI Act (4.2); the obligation to put in place policies to respect copyright law, especially regarding TDM opt-outs (4.3); the projected extraterritorial effect of such obligations (4.4); the transparency obligations (4.5); how the AI Act envisions compliance with such obligations (4.6); and potential enforcement and remedies (4.7). Section 5 offers some concluding remarks, focusing on the inadequacy of the current regime to address one of its main concerns: the fair remuneration of authors and performers.