A Competition Policy Analysis of Copyright Protection in Generative AI

Abstract

The rise of artificial intelligence (AI) has sparked significant debate, particularly regarding the relationship between generative AI (GenAI) and copyright. Indeed, GenAI appears to challenge every layer of copyright protection. Our analysis focuses on the tensions surrounding the use of copyrighted works to train AI models. Since AI training relies on vast amounts of data, two conflicting interests emerge. On one hand, copyright can act as a major barrier to entry, potentially stifling the next wave of technological innovation. On the other hand, GenAI systems may pose an existential threat to creative industries by replicating human creativity and producing literary and artistic works faster and at lower costs. Against this backdrop, policymakers worldwide are striving to balance these seemingly opposing interests. While most discussions focus on why and how copyright holders should be compensated, this paper examines when compensation is appropriate. To this end, it advocates for a competition-based approach in assessing the application of copyright limitations and exceptions. Specifically, it argues that antitrust tools can help courts and policymakers determine when creators suffer commercial harm and when AI-generated content may be considered a substitute for human creations.