Training Data and Copyright: Legal Boundaries and Economic Impact in AI Development

This dissertation aims to provide a comprehensive analysis of the current legal frameworks on generative AI tools in the field of international copyright, employing a comparative and qualitative approach to examine the United States and the European legal scenario. Starting with a technical explanation of the operational mechanisms of generative artificial intelligence systems and models, the study navigates the different legal and policy approaches adopted in industry-centered U.S. and rights-balance EU. Accounting for the numerous economics impacts stemming from the legal approaches adopted by these leading jurisdictions and global powers, this dissertation also touches upon questions of authorship and remuneration arising from generative AI outputs. Rather than seeking to establish a definitive resolution or singular approach, the research aims to objectively report on the existing frameworks, their efficacy, and the anticipated outcomes they may yield. Key questions that will guide this paper include: Are the current laws suited to balance authors’ rights against the technological development of Deep Learning technologies and their ever-evolving nature? What are wider and longer-term impacts on social and economic fields? What are future perspectives on flesh-and-blood authorship and what are safety measures that should – or must – be taken to preserve the right equilibrium for the common good and welfare? By analyzing different scholarly perspectives and recommendations, this dissertation endeavors to identify the most viable solutions for navigating the complexities of copyright law in the age of generative AI.