人工智能生成内容的著作权侵权风险与侵权责任分配  

Risk of Copyright Infringement and Distribution of Tort Liability for the Artificial Intelligence Generated Content

作　　者：刘云开 LIU Yunkai(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：《西安交通大学学报（社会科学版）》2024年第6期166-177,共12页Journal of Xi'an Jiaotong University：Social Sciences

基　　金：国家社会科学基金重大项目(21&ZD203);国家社会科学基金一般项目(20BFX134)。

摘　　要：在现行著作权法语境下,ChatGPT等生成式人工智能产品的产业化应用,将引发规模更大、隐蔽性更强的人工智能生成内容(AIGC)著作权侵权风险。算法比对技术的嵌入虽有助于降低侵权风险,但在侵权内容实质性相似判断中仍将面临技术失灵。由于AIGC服务提供者与网络用户在“算法创作—内容传播”链条中的角色发生置换,因此相较传统网络著作权侵权,AIGC著作权侵权主体的认定模式有所转变。在AIGC著作权侵权责任认定中,网络用户作为AIGC加工承揽合同中的定作人,一般不具有输出侵权内容的主观过错。AIGC服务提供者兼具侵权风险制造者和侵权内容提供者的角色,应被认定为直接侵权主体,不应以技术中立者的身份适用“避风港规则”。在AIGC著作权侵权责任的分配上,AIGC服务提供者承担无过错责任较之过错责任更为适宜;网络用户在传播侵权内容时存在过错的,应当承担损害赔偿责任,其免责事由可参照适用“避风港规则”。为实现著作权保护和产业创新的协同目标,可遵循责任规则启动AIGC著作权停止侵害请求权限制机制,并考虑引入第三方风险分担机制,构建AIGC著作权责任保险制度和赔付基金。The industrial application of ChatGPT and other generated artificial intelligence products will cause a larger and more hidden risk of AIGC copyright infringement. AIGC copyright infringement is different from the traditional network copyright infringement type, it is necessary to investigate its infringement risk, infringement subject and tort liability involved separately. Most researchers focuse on the copyright problem in machine learning, the copyrightability and ownership of AIGC, but there is some insufficient research on the copyright infringement of AIGC.This paper combines the operation principle of artificial intelligence technology and the basic principle of copyright law, and explores the causes of the risk of AIGC copyright infringement. Using the legal relationship analysis method, the legal relationship between artificial intelligence product designers, AIGC service providers and network users is clarified, and the subject of AIGC copyright infringement liability is locked. On this basis,this paper discusses how to distribute the copyright infringement liability between the above subjects.This paper believes that, due to technical reasons and legal provisions, the risk of AIGC copyright infringement is objective and difficult to eliminate. Although AIGC copyright infringement belongs to a form of network copyright infringement,due to the replacement of AIGC service providers and network users in the chain of “algorithmic creation—content dissemination”, compared with the traditional network copyright infringement, the identification mode of AIGC copyright infringement liability subject has changed. In terms of the specific distribution of tort liability, it is more appropriate for AIGC service provider to assume no-fault liability than fault liability;network users who are at fault when spreading the infringing content shall be liable for damages, but they can claim the exemption by referring to the safe harbor rules. In order to balance the goal of copyright protection and industrial innova

关 键 词：人工智能生成内容 生成式人工智能 著作权侵权 归责原则 风险分担机制 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] D923.41[自动化与计算机技术—控制科学与工程]