BALANCING FUNDAMENTAL RIGHTS, INNOVATION AND THE PUBLIC INTEREST IN COPYRIGHT LAW: THE CASE OF GENERATIVE AI MODELS

This thesis argues that generative artificial intelligence exposes a fundamental tension within the human-centric copyright framework, which is built around the anthropocentric standard of the “author’s own intellectual creation.” Under this framework, autonomously generated outputs are ineligible for protection as copyright requires a human creator capable of making ‘free and creative choices. The legal uncertainty emerges bot at the output and the input phase of AI. Notably, the thesis demonstrates that the text and data mining exceptions under the Copyright in the Digital Single Market Directive (CDSMD) struggle to accommodate the data-driven logic of machine learning. The conflict between ‘lawful source’ and ‘lawful conduct’ of the access requirement, alongside the shortcomings of the opt-out mechanism, reveals a constitutional challenge. By situating the EU framework in a comparative perspective with the UK and US, the thesis shows that instruments such as the AI Act introduce procedural transparency obligations yet avoid addressing the substantive copyright questions of authorship, reproduction and most importantly, the absence of a harmonised remuneration moder for authors whose works fuel AI development. This thesis concludes that addressing these tensions does not require abandoning copyright’s anthropocentric foundations but reinforcing them. The path forward is therefore to conceptualise AI as an advanced tool that supports but never substitutes the human author. The analysis reveals the need to clarify the law, particularly by considering statutory remuneration systems that ensure fair and sustainable access to training data. The thesis also points out the limits of individual enforcement and calls for a shift toward collective licensing mechanisms.