The virtual void : copyright challenges faced in the generation of AI works

Abstract

We live in an era where machines, having been fed with millions of pieces of human expression, are leveraging the extensive data they have collected to generate new, unconventional forms of art. While this has certainly been an entertaining endeavour for tech enthusiasts and a lucra:ve opportunity for AI developers, it has thus brought to the forefront important questions as to the implications for intellectual property rights in this creative process. Over the years, text and data mining (hereinafter: TDM) has solidified its position as a building block for Artificial intelligence (hereinafter: AI) companies, particularly those specialising in creating machine-generated artworks. Nevertheless, ‘the copyright black hole’ associated with the generation of AI art, has sparked heated controversies over the use of copyrighted material for algorithmic training, often acquired, without proper authorisation. To address these concerns the European Union (hereinafter: EU) introduced two mandatory exceptions for TDM activities, as outlined in Articles 3 and 4 in the Directive on Copyright in the Digital Single Market (hereinafter: DSM Directive). The thesis investigates how Article 4 DSM Direc:ve navigates the legal complexities involved in the use of authorship work in TDM activities, specifically in instances where such materials are used without the explicit consent of rightholders. It begins by examining the essence of Article 4 DSM Directive and the underlying concept of TDM. The analysis then continues by critically assessing whether the copyright issues related to TDM, were adequately resolved within the said Article. Finally, the thesis concludes that, despite the inclusion of Article 4 DSM Directive, significant shortcomings persist. It concludes by proposing further initiatives to better align the EU legal framework with the evolving demands of AI technology.