Copyright implications of the use of generative AI

Abstract

This thesis will be divided into three chapters: in the first chapter, I will talk about the copyright implications of training AI, which I refer to as inputs. To that end, I will acquaint the reader, on the one hand, with the statutory provisions concerning the temporary reproduction in article 5(1) of the InfoSoc Directive and the TDM exceptions in articles 3 and 4 of the DSM Directive. On the other hand, I will introduce the U.S. fair use doctrine and walk through how that would apply to Generative AI. In the second chapter, I will talk about the copyright implications of the AI -generated works, which I refer to as outputs. Since the EU has not armonized the infringement on the right of reproduction and adaptation (also known as derivative right), I will focus instead on the U.S. alone, since it is also the country hosting most AI companies. To this end, I will delve into the right of reproduction and, by extension, the right of derivatives. Finally, I will go through the fair use arguments from the perspective of the outputs. In the third and last chapter, I will ponder whether AI can be authors for the purposes of Copyright Law. If they cannot because they are not human, I will determine which, either the AI owner or the user, may own the copyright (if any) arising from a given AI generated work. To achieve this, I will go through how the situation is settled in the US, the EU and the UK.