The contrast between “strong” European copyright and U.S. copyright—which is weakened in particular by fair use—may have become too simplistic to describe the current developments in the field of AI. If U.S. courts continue to narrow the scope of fair use while the European Union fails to strengthen the evidence available to rights holders, the paradox could be that European law becomes less protective of authors than U.S. law. For several years, the European debate on artificial intelligence has revolved around a widely accepted assumption: unlike the United States, the European Union has a copyright system that is, by definition, more favorable to creators, and consequently, European creators would be better off when dealing with AI developers than their American counterparts. However, recent developments in U.S. case law and the practical difficulties in applying the European framework call for a more nuanced interpretation. The first rulings handed down in U.S. litigation pitting rights holders against publishers of artificial intelligence models do not, in fact, establish a fair use applicable to all training activities. There will therefore be no general authorization to use works without permission or compensation, contrary to what many AI developers had hoped. Instead, U.S. judges are beginning to examine the various stages of the technical process, as well as the origin of the content…
