AI covers: legal notes on audio mining and voice cloning

Abstract This article explores the impact of Artificial Intelligence (AI) on the music industry, particularly focusing on the case of AI-generated covers. The emergence of AI technologies has been raising concerns not just about the originality and protection of AI-generated outputs but also about the complex input and training phase of those systems. The focus of this contribution is the latter, analysing the case of AI covers from the perspective of copyright and image rights. In the first part, an overview of the text and data mining (TDM) exception found in Article 4 of Directive 2019/790 is presented, with a primary focus on the opt-out mechanism in connection with the three-step test. Moving to the second part, the analysis delves into the complexities of voice cloning, highlighting the absence of a comprehensive European Union regime for image rights. By addressing these issues, this contribution unveiled two crucial points. First, AI models trained on various artists’ works to create and spread deepfake covers not only violate copyright but also reveal shortcomings in the TDM exception. Second, while the multifaceted image right regime may not be as exhaustive as necessary, it proves to be a viable solution against voice cloning with anticipated advancements in the future.