人工智能生成内容的著作权归属问题研究

Research on the Copyright Ownership of Artificial Intelligence-Generated Content

摘要: 随着生成式人工智能技术的快速发展，人工智能生成内容对以“人类作者”为核心的著作权制度提出了深刻挑战，本文围绕人工智能生成内容的著作权归属问题，系统梳理了可版权性的理论分歧，分析了否定说与肯定说的法理依据，并在此基础上提出以“人类实质性贡献”作为判断人工智能生成内容可版权性的标准。进一步从使用者、设计者、训练者以及投资者等多元主体出发，讨论了生成内容的权利归属问题以及提出了构建类型化权利归属的规则。对于由人工智能完全自主生成且缺乏人类实质性干预的内容则不能构成作品，应进入公有领域；而人机协作生成的内容，有约定从约定，无约定时则推定使用者为著作权人；在职务或委托场景下，优先适用现行著作权法相关规定。此外，文章还探讨了上游训练数据版权问题、邻接权路径的可行性、标识义务的衔接以及设置独创性门槛与保护期限制等制度配套措施，旨在为人工智能生成内容的著作权治理提供系统性规范路径。

Abstract: With the rapid advancement of generative artificial intelligence technology, AI-generated content has posed profound challenges to copyright systems centered on “human authorship”. This article systematically examines the theoretical disputes regarding the copyright ownership of AI-generated content, analyzes the legal justifications for both the negative and affirmative positions, and proposes that “substantial human contribution” should serve as the criterion for determining copyrightability. It further explores the allocation of legal rights among multiple stakeholders—including users, designers, trainers, and investors—and suggests establishing a typified framework for rights attribution. Content entirely generated by AI without substantial human involvement does not qualify as a work and should enter the public domain; in cases involving human-AI collaboration, copyright shall be governed by contractual agreements where applicable, or otherwise presumed to belong to the user. In employment or commissioned contexts, existing copyright law provisions shall take precedence. Additionally, the article discusses related issues such as copyright in upstream training data, the feasibility of adjacent rights approaches, the integration of identification obligations, and complementary measures including setting originality thresholds and limiting protection periods, aiming to provide a systematic regulatory framework for governing copyright in AI-generated content.