Abstract This article is a ‘metaphorical’ guide to today’s most pressing artificial intelligence (AI) copyright questions, focusing in particular on the EU and the USA. Is unauthorized training on copyright-protected works permitted? Can AI models copy? And is AI-generated output itself protected? As this article demonstrates, debates on these questions can all be traced back to a handful of crucial metaphors. After all, generative AI is hardly comprehensible without the extensive use of metaphors and analogies. Most notably, AI is systematically conceptualized in human terms such as ‘neural networks’ that ‘learn’, ‘know’ or ‘memorize’. This article aims to demonstrate how such metaphors (unconsciously) influence legal evaluations and even judicial decisions in copyright law. The resulting analysis is particularly relevant to lawyers, judges and artists interested in copyright and its intersection with AI. Yet, it may also appeal to those interested in AI, legal reasoning and language more generally, as metaphors and their (rhetorical) effects are by no means unique to copyright and may be equally relevant in fields such as privacy law and (legal) philosophy.
Michiel A Smit (Fri,) studied this question.