The European Parliament's March 10 resolution on AI and copyright is non-binding, but its significance should not be underestimated. The proposal to extend EU copyright territoriality to any AI system offered in the European market — regardless of where the underlying model was trained — would, if enacted, represent one of the most consequential expansions of intellectual property jurisdiction in the digital era. The logic is straightforward: an AI model trained in California or Singapore but deployed to European users should be subject to European copyright obligations, just as a physical product manufactured abroad must comply with EU consumer protection law when sold within the single market. The 5–7% of global turnover levy proposed as retroactive compensation is the figure that will generate the most controversy. Applied to the revenues of OpenAI, Google DeepMind, or Anthropic, it would represent sums in the billions — a number that makes the individual licensing deals struck with Condé Nast or the Associated Press look like rounding errors. The European Writers Council's analysis of the resolution is instructive: the clarification of territoriality is framed not as a punitive measure but as a level-playing-field mechanism, ensuring that European and non-European AI providers compete on equal terms within the EU market. Whether the Commission will translate this resolution into binding legislation — and on what timetable — remains to be seen. But for publishers and rights holders operating across the Atlantic, the direction of travel is unmistakable.