Imprese innovative e prodotti delle intelligenze artificiali generative (IAG): opportunità e limiti nelle regole del diritto d’autore

Generative Artificial Intelligence raises significant challenges in copyright law, particularly regarding the use of creative works for the training of Large Language Models (LLMs). The EU’s AI Act addresses the matter with two obligations towards providers of general-purpose AI models: on the one hand, the reference (Art. 53.1(c)) to the reservation of rights tied to Text and Data Mining (Art. 4.3 of Directive (EU) 2019/790) both discourages AI-driven uses and does not fully protect rightsholders’ interests, as the opt-out provision still appears to be unclear on many legal and technological aspects; on the other hand, the disclosure obligation of the training content (Art. 53.1(d)) might be a step in the right direction, as it provides some light on the actual utilization of works. However, since computational analysis, following lawful access to content, does not clearly fall under acts of reproduction in the EU regime, a balanced solution might come through the introduction of paid-but-permitted measures. This framework would not only foster innovation in AI development, but it would also ensure adequate remuneration for authors and creators.