The originality test in works created during the era of generative AI: the role of instructions (prompts) in copyright and the fate of the “old school” author

Abstract

Generative Artificial Intelligence (AI) has important implications for copyright law. In particular, the question has arisen to what extent the use of prompts can give rise to protection for the output of AI tools. Taking the criteria for copyright protection in the European Union and the United States of America as starting points, this article analyzes decided cases in both jurisdictions, as well as a pilot case in China, and predicts that, in the near future, courts are likely to attribute copyright protection to natural persons in works created through the use of prompts, on a case-by-case basis. Generative AI makes it impossible to determine whether a work, on its face, is sufficiently human-authored to be protected by copyright. Given the risk of strategic behavior by private parties who do not disclose the use of AI or who rewrite prompts ex post, judges will necessarily have to move away from the implicit assumption that what appears creative, must have had a human author who made creative choices. It is likely that (alleged) authors will increasingly be required to prove that they made such choices. This, in turn, risks raising the burden of proof for old-school authors who do not use AI.