DefinePK

Abstract

Generative artificial intelligence (AI) is rewriting the relationship between creators, rights holders and developers. Its capacity to synthesize text, images and sound from vast troves of online data has triggered a wave of copyright disputes, while regulators scramble to catch up. Article 53(1)(d) of the European Union (EU) AI Act introduces a novel requirement: providers of general‑purpose AI (GPAI) must publish a “sufficiently detailed summary” of the training data used to develop their models. Proponents claim that forcing transparency will shift disputes from litigation to licensing. Critics warn that generic summaries will not give rights holders enough information to assert their rights and that trade secrets may be jeopardized. This paper asks whether Article 53(1)(d)’s training‑data summary measurably shifts developer behavior toward licensing over litigation and alters rightsholder enforcement across jurisdictions. We review the evolving legal landscape, synthesize empirical and theoretical scholarship, and analyze trends in licensing deals and lawsuits before and after the adoption of Article 53(1)(d). Our findings suggest that transparency obligations are necessary but insufficient: while high‑impact lawsuits remain, the availability of documented training‑data summaries correlates with an uptick in licensing agreements, particularly in the EU. Trade‑secret tensions, uneven enforcement and divergent fair‑use doctrines continue to complicate the path toward a balanced global framework. We conclude with recommendations for refining disclosure templates, harmonizing opt‑out mechanisms and calibrating economic incentives to ensure that generative AI innovation and creative rights coexist.