AUTHORSHIP WITHOUT AUTHORS? RETHINKING COPYRIGHT OWNERSHIP IN GENERATIVE AI ACROSS JURISDICTIONAL BOUNDARIES

Resumo

This article argues that the most acute copyright problem posed by generative AI is not simply whether “AI outputs” can be protected, but how ownership rules behave once authorship is destabilised across borders. Current doctrine assumes a human author whose creative agency anchors originality, moral rights, initial ownership, and term. Generative AI disrupts that architecture by separating training-stage reproduction of protected works from output-stage production of new expressive material whose “author” is often indeterminate, distributed, or strategically framed by platform contracts. Comparative analysis indicates a converging human-centred baseline in the United States and European Union, while the United Kingdom retains a statutory attribution rule for computer-generated works that sits uneasily with modern originality doctrine and international harmonisation pressures. This article argues that extending full copyright to autonomously generated outputs would amplify legal fragmentation and power asymmetries, creating extraction-friendly defaults where proprietary model operators or sophisticated users capture value from global cultural inputs. The article proposes a doctrinal solution built around a trans-jurisdictional “human creative control” threshold, complemented by provenance-oriented disclosure and licensing mechanisms focused on training.