生成式人工智能产品责任认定研究

生成式人工智能产品在责任认定上存在特殊性，选择过错责任而非严格责任作为生成式人工智能产品责任的主要归责原则，有助于实现“预防风险、鼓励创新”的价值目标。对人工智能产品技术支持者的责任认定，可以类推适用产品责任中的现有技术抗辩，即以“现有技术水平”作为认定其是否存在过错的标准，而“现有技术水平”的具体内涵需要结合时间和行业两个维度综合考量。对人工智能产品服务提供者的责任认定，不宜类推适用避风港规则，可以类推适用公共场所、经营场所管理者所承担的安全保障义务，也就是以安全保障义务的有无和履行情况判断其过错。

Research on attribution of responsibility for generative artificial intelligence products

Generative artificial intelligence（AI） products have special characteristics in liability determination. Choosing fault liability instead of strict liability as the main attribution principle for generative AI products can help achieve the value goal of “risk prevention and innovation encouragement”.The responsibility determination of technology supporters of AI product can be applied by analogy to the existing technology defense in product liability， which uses “existing technology level” as the standard to determine whether there is fault， and the specific connotation of “existing technology level” needs to be comprehensively considered from two dimensions： time and industry. It is not advisable to analogously apply the Safe Harbor Privacy Principles to AI product service providers. The safety obligations undertaken by managers of public places and business premises can be analogously applied to service providers， judging the fault of service providers based on the existence and fulfillment of safety obligations.