生成式人工智能的责任认定研究  

Research on Responsibility Attribution of Generative Artificial Intelligence

作　　者：孙文波 SUN Wenbo(School of Law,University of Chinese Academy of Social Sciences,Beijing 102400,China)

机构地区：[1]中国社会科学院大学法学院,北京102400

出　　处：《唐山学院学报》2025年第2期70-79,共10页Journal of Tangshan University

摘　　要：生成式人工智能的多场景适用正呈现出不可阻挡的嬗变趋势,而在其实现规模化应用的同时却伴生着多维风险,风险“进化”的损害则又会引致归责问题。以往归责路径中,人工智能的侵权责任由提供者、管理者或使用者承担。如今,生成式人工智能存在成为法律拟制主体的可能,因此其责任认定应有别于传统认知论。建议以赋予生成式人工智能有限法律人格为逻辑起点、以围绕“生命周期”分阶段认定责任为逻辑顺延、以处财产责任抑或增添责任类型为逻辑闭环,来促进人工智能从“可控制”发展至“可规制”。The multi-scenario application of generative artificial intelligence(AI)is showing an unstoppable trend of transformation.However,while achieving large-scale application,it is accompanied by multi-dimensional risks,and the“evolution”of these risks leads to issues of responsibility attribution.In previous attribution paths,the tort liability of AI was borne by providers,administrator or users.Now,generative AI has the potential to become a legal person,so its responsibility attribution should differ from traditional cognitive theories.It is suggested to start with the logic of granting generative AI limited legal personality,extend the logic of phased responsibility attribution around the“life cycle”,and close the logic loop with property liability or the addition of liability types,thereby promoting the development of AI from“controllable”to“regulatable”.

关 键 词：生成式人工智能 ChatGPT 有限法律人格 财产责任 辅助第三人 

分 类 号：D923[政治法律—民商法学]