A copyright analysis of infringement liability in the course of Artificial Intelligence training Process

Abstract

Artificial Intelligence (AI) systems have radically changed traditional definitions of human creativity. AI is not a mere computer program; rather, it has become a talented creator and author in and of itself operating with limited human involvement (or entirely independent of it). This dissertation intends to explore how current legal frameworks might govern potential copyright infringement liabilities regarding data inputted in the course of training AI applications. With the abilities vested in Machine Learning, technology can imitate human creativity by drawing from vast datasets, surpassing them in creating novel works. The fact that the workflow of AI often entails using copyrighted work poses concerns about the legal legitimacy of the training process. However, an exploration of possible infringement liabilities in the training process must be pre-empted by presenting the reader with an overview of the interaction between Artificial Intelligence and copyright law, imparting an understanding of technical aspects related to generating creative works. In the second section of this dissertation, prior to delving into the potential infringements in the training data, relevant copyright provisions on a national and international scale must be to be laid down in order to specify which restricted acts may be violated by the training process. The primary focus of this study is on what may amount to copyright infringement in training processes, and moreover, how national courts would react in conventional copyright litigation. In order to provide a fully and thoroughly answer, I will be enacting a comparative legal analysis of United Kingdom and European copyright standards through scrutinising case law, looking for interpretations that might illuminate the answer. Before finishing, I intend to put forward a question as to whether AI workflows are exempted by law. Therefore, I aim to explore two relevant exceptions which seem fitting and applicable