高校图书馆工作

2025, v.45;No.226(02) 6-12

[打印本页] [关闭]
本期目录(Current Issue) | 过刊浏览(Archive) | 高级检索(Advanced Search)

生成式人工智能训练数据使用的著作权困境及其破解
The Copyright Dilemma in the Use of Training Data for Generative Artificial Intelligence and Its Resolution

谢星辰,宋尧,李可心
Xie Xingchen,Song Yao,Li Kexin

摘要(Abstract):

生成式人工智能训练数据作为技术创新的基础性资源,其合规使用对推动算法优化与产业迭代具有战略意义。然而,传统著作权框架下的授权使用、合理使用、法定许可等规则已捉襟见肘,生成式人工智能的海量数据需求与现行著作权制度形成冲突,并演变为制约人工智能产业创新的法律桎梏。文章通过规范分析与比较,详细阐述了生成式人工智能训练数据使用的著作权困境及原因;基于对美欧日制度实践的批判性考察,提出建构我国生成式人工智能训练数据著作权例外制度的三重路径:一是重构合理使用规则,将“信息分析型使用”纳入豁免范围并确立“无市场冲突”判断标准;二是创新准法定许可制度,通过“公告+异议排除”机制建立弹性授权路径;三是探索著作权集体管理组织路径,构建“默认许可+精准分润”的规模化授权体系。以消解权利保护与产业发展之间的矛盾,避免制度遏制创新,防止创新侵蚀权利。
Training data for generative artificial intelligence(AI) functions as a foundational resource for technological innovation, and its lawful utilization holds strategic significance for algorithmic advancement and industrial transformation. However, traditional copyright frameworks—centered on authorized use, fair use, and statutory licensing—are increasingly inadequate. The enormous data demands of generative AI conflict with existing copyright regimes, resulting in legal constraints that hinder AI-driven innovation. Through normative and comparative analysis, this paper examines in detail the copyright challenges associated with using training data in generative AI and the structural causes underlying these issues. Drawing upon a critical review of practices in the United States, European Union, and Japan, the study proposes a tripartite approach to building a copyright exception framework for generative AI training data in China: first, restructuring fair use by including “analytical use of information” within its scope and establishing a “no market harm”criterion; second, developing a quasi-statutory licensing system through a “public notice plus objection exclusion” mechanism to enable flexible authorization; and third, exploring a collective management approach to establish a scalable system based on “ default licensing plus precise revenue sharing. ” These proposals aim to reconcile the tension between rights protection and industrial development, mitigating the risk of regulatory suppression of innovation while safeguarding copyright interests.

关键词(KeyWords): 生成式人工智能;训练数据;授权使用;合理使用;法定许可
Generative artificial intelligence;Training data;Authorized use;Fair use;Statutory license

Abstract:

Keywords:

基金项目(Foundation): 国家社会科学基金重大项目“总体国家安全观下产业知识产权风险治理现代化研究”(项目编号:21&ZD204)的研究成果之一

作者(Author): 谢星辰,宋尧,李可心
Xie Xingchen,Song Yao,Li Kexin

参考文献(References):

扩展功能
本文信息
服务与反馈
本文关键词相关文章
本文作者相关文章
中国知网
分享