Copyright Ownership of AI Generated Content: Evidence from Judicial Practice and a User Attribution Rule

Abstract

Ensuring a clear and workable allocation of copyright ownership for AI generated content has become one of the most urgent challenges of contemporary copyright law. This article examines how existing copyright doctrines, especially originality and authorship, respond to outputs produced with generative artificial intelligence, and how ownership should be attributed in a way that preserves the human centred foundations of copyright while maintaining effective incentives for creation and dissemination. The study adopts doctrinal analysis and comparative legal research, with particular focus on Chinese judicial practice and United States copyright registration practice. It further integrates a practice based review of representative cases, including Tencent Dreamwriter, the Spring Breeze image dispute, and Beijing Feilin v Baidu, together with United States determinations such as Zarya of the Dawn, Théâtre D’opéra Spatial, and A Single Piece of American Cheese. The findings indicate broad convergence on the principle that artificial intelligence is not a legal subject and cannot be an author, while jurisdictions diverge on the threshold and proof of human creative control. Building on these results, the article argues for a user attribution approach as the general rule, supported by clearer contribution assessment standards, contractual flexibility as a limited exception, and a workable labelling mechanism to enhance transparency and reduce disputes.