Models of Ownership for Works Created by Artificial Intelligence: A Comparative Study with Emphasis on Iranian Law

The evolution of artificial intelligence from early symbolic processing paradigms to contemporary generative systems has posed a fundamental question for intellectual property discourse: In a legal framework, how should works lacking direct human creative intervention be analyzed? In existing literature, two predominant conceptions of AI's nature can be observed: one group, denying its intelligence, emphasizes social consequences and the necessity of control, while the majority of experts, acknowledging the tangible reality of this technology's outputs, have pursued the issue within the framework of legal regulation. Nevertheless, determining the shares of the programmer, the system owner, and the user in generating outputs is central to ownership debates. The three prevalent theories—anthropocentric, attribution-to-the-machine, and entry into the public domain—each face significant limitations in explaining originality, liability, and economic incentives. This is because the structure of intellectual property systems is founded on creating a balance between private rights and public interests, and the complete removal of protection could weaken economic incentives and investment in technological development. Furthermore, the public domain theory does not address the core question of the "work's capacity for ownership" and has limited functional efficacy. Comparative analysis also indicates a growing trend in legal systems toward hybrid and multi-layered models. On this basis, the proposed view is that "ownership based on predominant influence" could provide a balanced and effective response to generative AI works by offering precise criteria for measuring the roles of both human and system.