生成式人工智能机器学习的版权分层规制模式——以“表达性使用”为视角  

The Hierarchical Regulation Model for Copyright in Generative AI Machine Learning:From the Perspective of"Expressive Use"

作　　者：王诗童 杨利华[1] WANG Shi-tong;YANG Li-hua(Civil,Commercial and Economic Law School,China University of Political Science and Law,Beijing 100088,China)

机构地区：[1]中国政法大学民商经济法学院,北京100088

出　　处：《编辑之友》2025年第2期79-86,共8页Editorial Friend

基　　金：北京市习近平新时代中国特色社会主义思想研究中心重点项目“习近平法治思想中知识产权保护重要论述研究:理论体系与实践运用”(22LLFXB038-2022)。

摘　　要：受版权法保护的文学、艺术和科学作品是生成式人工智能机器学习的高质量语料资源。为促进生成式人工智能技术与相关行业的发展,需合理界定生成式人工智能机器学习使用作品的版权保护边界。无论是人类中心主义的价值内核、推动文化创新的立法宗旨,还是作品利益调整的制度功能,都体现出版权法内在的“表达性使用”理论预设。文章认为基于过程视角下使用作品的性质、指向与方式,以及结果视角下使用作品的效果,一般意义上的生成式人工智能机器学习对作品的使用是一种区别于“表达性使用”的非表达数据性使用,不在版权法的保护范围之内。“表达性使用”视角下的生成式人工智能机器学习应设置分层分级的版权规制路径,以促进生成式人工智能产业的发展。Literary,artistic,and scientific works protected by copyright law constitute high-quality corpus resources for generative AI machine learning.To facilitate the development of generative AI technology and related industries,it is necessary to reasonably define the copyright protection boundaries for the use of works in generative AI machine learning.Whether it is the human-centric value core,the legislative purpose of promoting cultural innovation,or the institutional function of adjusting the interests of works,they all reflect the inherent theoretical presumption of"expressive use"under the jurisdiction of copyright law.Based on the nature,direction,and manner of using works from a process perspective,as well as the effects of using works from a result perspective,the use of works in generative AI machine learning in a general sense is a non-expressive,data-oriented use distinct from"expressive use"and falls outside the scope of copyright protection.For generative AI machine learning from the perspective of"expressive use",a hierarchical copyright regulation pathway should be established to promote the development of the generative Al industry.

关 键 词：生成式人工智能 机器学习 版权法 “表达性使用” 合理使用 

分 类 号：D923[政治法律—民商法学]