Experimental and Behavioural Economics Seminar Series Joint with Economic Theory - Jay Lu (UCLA)

Title: Creative Ownership in the Age of AI

Abstract: Copyright law asks whether a new work is “substantially similar” to an existing one, but generative AI can closely imitate a creator’s style without copying any content—a capability now central to ongoing litigation. We argue that existing definitions of creative ownership are ill-suited to this setting and propose a new criterion: an AI-generated output depends essentially on an existing work if it could not have been generated without that work in the training corpus, and it is permissible if it does not depend essentially on any existing work. We establish structural properties of the set of permissible outputs and prove a sharp asymptotic dichotomy: when the creative process is light-tailed, dependence on individual works eventually vanishes, so in the long run the criterion places no limits on AI generation; on the other hand, when it is heavy-tailed, individual works remain pivotal and the criterion continues to bind even for arbitrarily large corpora.