Generative artificial intelligence outputs, copyright infringement, and the assignment of liability

Generative artificial intelligence outputs, copyright infringement, and the assignment of liability

연구 분야: Safety

학회: AI and Ethics

As it stands today, the field of copyright litigation is flush with generative artificial intelligence (AI) cases. Owing to AI’s novelty, the arguments being advanced in these cases are untested, and their effectiveness remains uncertain. This article will explore a subset of the claims made in such cases: claims of copyright infringement stemming from AI-generated materials. Its first section will outline how liability for copyright infringement is established, finding that at least some generative AI outputs are copyright-infringing and that this fact poses a major threat to the continued existence and development of generative AI models. Its second section will contend that, under current copyright law, generative AI developers can be held vicariously and contributorily liable for infringing AI outputs, and that generative AI users can be held directly liable for infringing AI outputs. Its third section will advance a normative argument for this dual assignment of liability on the basis of both natural law and utilitarian theories of copyright.