生成式人工智能侵权责任的分层配置研究——基于责任主体与归责原则的协同视角

A Study on the Layered Allocation of Tort Liability for Generative Artificial Intelligence—From the Synergistic Perspective of Liability Subjects and Imputation Principles

摘要: 生成式人工智能技术的快速发展推动了内容生产方式的深刻变革，同时也引发了著作权侵权、人格权侵害、个人信息泄露及虚假内容传播等多元侵权风险。传统侵权责任制度以行为人实施侵权行为为前提构建责任体系，而生成式人工智能具有自主生成、技术链条复杂、主体协同参与以及风险来源多元等特点，使开发者、模型提供者、服务提供者、平台运营者及用户共同参与内容生成过程，传统以单一主体为中心的侵权责任体系面临适用困境。当前学界围绕生成式人工智能侵权责任形成了统一归责、无过错责任、类型化归责以及分层责任等不同理论路径，但现有研究多聚焦于归责原则、证明责任或特定主体责任配置，缺乏责任主体划分、归责原则选择与证明责任配置之间的体系化协调。基于此，本文从技术运行流程与风险形成机制出发，以责任主体分层为基础，以归责原则协同适用为核心，探讨兼顾技术创新与权益保护的生成式人工智能侵权责任分层配置体系，以期为我国生成式人工智能侵权责任制度的完善提供理论参考。

Abstract: The rapid advancement of generative artificial intelligence technology has profoundly transformed modes of content production, while simultaneously giving rise to diverse tort risks such as copyright infringement, infringement of personality rights, disclosure of personal information, and dissemination of false content. Traditional tort liability systems construct liability frameworks on the premise that an actor commits a tortious act. However, generative AI is characterized by autonomous generation, complex technological chains, collaborative participation of multiple subjects, and diverse risk sources, involving developers, model providers, service providers, platform operators, and users in the content generation process. Consequently, the traditional tort liability regime centered on a single subject faces difficulties in application. The current academic community has developed various theoretical approaches to tort liability for generative AI, including unified imputation, strict liability (liability without fault), typified imputation, and layered liability. Nevertheless, existing research predominantly focuses on imputation principles, the burden of proof, or the allocation of liability to specific subjects, lacking systematic coordination among the division of liability subjects, the choice of imputation principles, and the allocation of the burden of proof. On this basis, proceeding from the technical operation process and risk formation mechanism, this paper takes the stratification of liability subjects as the foundation and the synergistic application of imputation principles as the core, and explores a layered allocation system of tort liability for generative AI that balances technological innovation and the protection of rights and interests, with a view to providing theoretical reference for the improvement of China’s tort liability regime for generative artificial intelligence.