摘要： 受技术复杂性及法律滞后性的双重影响,生成式人工智能生成错误个人信息在侵权责任成立认定阶段存在保护权益范围不清、归责原则混乱、责任成立认定规范冲突及构成要件尤其是因果关系难以证明等问题。对此,应以实现“促进生成式人工智能健康发展和规范应用”与“保护公民合法权益”这两项基本目标的平衡为出发点,通过侵害的直接性标准等确定生成式人工智能生成错误个人信息侵权场合的权益保护范围;通过类型区分的方法明确归责原则,在涉及个人信息处理者侵权时为过错推定归责,其他的为一般过错归责;并在过错归责的基础上确定责任成立的规范基础,将《民法典》第1165条第1款等与第998条分别采用的构成要件与利益权衡两种方法在作用于生成式人工智能生成错误个人信息侵权责任成立的认定时结合起来,共同展开相应法律效果的评价工作。当难以确定谁的行为导致错误个人信息生成时,可以类推适用《民法典》第1254条第1款的补偿规则,由可能加害的生成式人工智能服务提供者与使用者补偿受害人,以在受害人保护与人工智能产业发展之间取得平衡。

关键词: 生成式人工智能, 错误个人信息, 归责原则, 过错推定, 构成要件

Abstract: Influenced by both technological complexity and legal lag, the determination of tort liability for erroneous personal information generated by artificial intelligence faces several challenges.These include unclear scope of protected rights and interests, confusion in imputation principles, conflicts in norms for establishing liability, and difficulties in proving constitutive elements, particularly causation.To address these issues, it is essential to start by balancing the two fundamental goals of “promoting the healthy development and standardized application of generative artificial intelligence” and “protecting the lawful rights and interests of citizens.” The scope of rights and interests in cases of infringement arising from AI-generated erroneous personal information should be defined by applying criteria such as the directness of the infringement.Through categorical differentiation, the imputation principles in such cases should be clarified: a presumption of fault imputation applies when personal information processors are involved in the infringement, while general fault imputation applies in other scenarios.On the basis of fault imputation, the normative foundation for establishing liability should be determined by integrating the two approaches—constitutive elements and interest balancing—adopted respectively in Article 1165, Paragraph 1, and Article 998 of the Civil Code.When applied to determining the establishment of tort liability for AI-generated erroneous personal information, these approaches should be combined to jointly evaluate the corresponding legal effects.When it is difficult to identify whose conduct led to the generation of erroneous personal information, the compensation rule in Article 1254, Paragraph 1 of the Civil Code may be applied by analogy.In such cases, compensation may be granted to the victim by potentially liable generative AI service providers and users, thereby balancing victim protection with the development of the AI industry.

Key words: Generative Artificial Intelligence, Incorrect Personal Information, Imputation Principles, Presumption of Fault, Constituent Elements