生成式人工智能服务提供者侵权归责原则之辨  

On the Attribution Principle of Tort Liability for Generative Artificial Intelligence Service Providers

作　　者：徐伟 Xu Wei

机构地区：[1]上海政法学院

出　　处：《法制与社会发展》2024年第3期190-204,共15页Law and Social Development

基　　金：上海市哲学社会科学规划课题“生成式人工智能服务提供者侵权责任研究”(2023BFX014)的阶段性成果。

摘　　要：生成式人工智能服务提供者侵权适用过错责任原则存在诸多困难,生成式人工智能服务提供者侵权与过错责任原则的基本理念不相契合,以“行业通常技术水平”或“违法推定过失”认定过错难以有效解决过错的证明难题。生成式人工智能服务提供者侵权采过错推定责任原则具有一定的合理性,符合服务提供者与受害人地位不平等、技术复杂性高、信息不对称的现状,符合应当强化受害人救济的现实需求。然而,认定服务提供者存在过错与服务提供者无法预见和控制致害内容生成的客观事实相悖,会冲击过错推定责任理论体系。生成式人工智能侵权与个人信息侵权不同,不宜参照适用后者的归责原则。生成式人工智能侵权作为一种新型危险责任类型,适用无过错责任原则有其正当性:服务提供者能够从服务提供行为中获益,有能力分散损失,无辜的受害人应受到优先保护,且无过错责任原则有助于降低人工智能风险、减少制度不确定性引发的成本,能够为服务提供者提供可预期的制度环境。The application of fault liability to the tort liability of generative AI service providers presents numerous challenges,including inconsistency with the basic ideas of fault liability,and the problem of proving fault in practice remains unresolved even when the fault is determined by“industry-standard technical level”or“negligence per se”.Although the application of presumed fault liability on the liability of generative AI service providers is rational to a certain extent,including the inequality between service providers and victims,the high complexity of technology,information asymmetry,and the need to strengthen the protection of victims,the determination that the service providers are at fault contradicts the fact that they cannot foresee or control the output of harmful content,and it impacts the coherence of the theoretical framework of presumed fault liability.The infringement by generative AI differs from the infringement upon personal information and should not refer to the latter's principle of attribution.The infringement by generative AI should be regarded as a new scenario of dangerous liability,and the no-fault liability should be applied.The reasons for nofault liability are as follows:the service providers benefit from their actions and can distribute damages,innocent victims should be prioritized for protection,and no-fault liability helps to reduce the risks of AI,decrease costs induced by uncertainty of rules,and provide a predictable institutional environment for service providers.

关 键 词：生成式人工智能 归责原则 可预见 危险责任 无过错责任原则 

分 类 号：D923[政治法律—民商法学] D922.17[政治法律—法学]