Metaphors we judge (AI) by: a rhetorical analysis of artificial copyright disputes
I. Introduction
Generative artificial intelligence (GenAI) is too complex to fathom directly in a technological sense. At the very least, that is true for most lawyers, who instead orbit these ‘black boxes’ metaphorically without ever touching their core. As a consequence, language and rhetoric matter even more than they did before in copyright disputes. This article aims to analyse how the key questions concerning copyright and AI, indeed, largely revolve around metaphorical thinking, and how these metaphors affect our legal reasoning and judgements on these matters. In doing so, it provides an overview of European and US approaches to the most essential AI-related copyright questions with a strong emphasis on language.
The first legal question to be addressed is whether unauthorized training on data is allowed under copyright. A debate in which the metaphor ‘GenAI learns’ plays a crucial role. Next, the ‘memorization’ question will be discussed: do AI models actually copy works on which they are trained, or do they merely memorize or remember? Finally, the matter of copyrightability: can AI-generated output be protected by copyright? As will become apparent, this debate hinges on the question whether GenAI is just a ‘tool’ or ‘instrument’ or something more. But before addressing these questions, it is important to begin with an introduction to metaphors and why these matter to lawyers, particularly those dealing with AI and its intersection with copyright.
