ABSTRACT
This chapter examines the two central copyright questions raised by artificial intelligence: whether AI-generated works can attract copyright protection, and whether training AI models on copyrighted material constitutes infringement. On authorship, it surveys developments across China, the UK, and the United States, finding an emerging consensus that purely autonomous AI outputs are not copyrightable, but works involving sufficient human creative intervention may qualify for protection, positioning generative AI as a tool akin to Photoshop rather than an independent author.
On infringement, the chapter distinguishes between the input phase (copying works into training datasets) and the output phase (models reproducing protected content). It analyses text and data mining exceptions under UK law, the EU’s DSM Directive, and US fair use, noting that no jurisdiction has yet reached a definitive resolution. The chapter argues that output litigation may ultimately prove more consequential, raising novel questions about secondary liability and the applicability of intermediary liability frameworks to AI providers. Finally, it examines the copyright-specific provisions of the EU AI Act, including GPAI transparency obligations and the Code of Practice, situating these within the broader effort to balance innovation with the protection of creators’ rights.
Guadamuz, Andres, Artificial Intelligence and Copyright (December 12, 2025), in Law, Policy, and the Internet, Edwards L, Goanta C and Urquhart L (eds), Bloomsbury Academic (2026).
Leave a Reply