Research on Copyright Infringement Issues in Generative Artificial Intelligence

Abstract

Generative Artificial Intelligence (GAI) technology has achieved rapid development since the release of ChatGPT and Sora, with significant breakthroughs in domestically developed models. However, the copyright legal issues it raises are becoming increasingly prominent. This paper, based on the principles of GAI generation and the originality standard of works, argues that GAI training data meeting the " originality " standard possesses the attributes of copyrightable material. China's current Copyright Law and related legal norms are insufficient to address practical problems such as disputes over the copyright status of GAI training data, the inapplicability of moral rights, ambiguity of rights holders, and difficulties in pursuing infringement liability. Drawing on international legislative experience, this paper proposes that China should clarify the protectability of GAI training data by adding copyright exception clauses, attributing its copyright to users while appropriately restricting moral rights. Simultaneously, it should distinguish the duty of care and infringement liability of platform developers and users, constructing a copyright regulatory system that aligns with the development of GAI technology, achieving a balance between rights protection and innovation in the literary and artistic fields, and promoting the sustainable development of generative artificial intelligence.