人工智能应用责任的主体识别与归责设计

Subject Identification and Responsibility Distribution of AI Application Liability

应用责任可以同时适用于人工智能产品和服务,全方位解决人工智能部署应用端的风险致害责任,是未来我国《人工智能法》需要重点关注的课题。在明确《人工智能法》三分主体架构的基础上,按照救济、一站式与风险原则,不宜将开发者作为人工智能应用责任的第一责任主体,至于提供者、使用者则需要区分人工智能产品和服务分别界定。整齐划一的归责思路缺乏灵活性,应当引入基于风险的类型化思维,不仅契合侵权责任归责的一般原理,也是贯彻人工智能分级治理的理念。对于高风险人工智能,确立无过错责任有助于充分救济受害人,公平分配人工智能技术的风险,避免归责体系的内在冲突;对于有限风险与低风险人工智能,应当分别适用过错推定责任与过错责任,以实现风险与归责的动态平衡。

Application liability can be applied to AI products and services at the same time,and comprehensively solve the harm liability of AI application,which is a topic that needs to be focused on in China's AI Law in the future.On the basis of clarifying the structure of the three-part subject of the AI Law,and in accordance with the principles of relief,one-stop and risk,it is not appropriate to regard developers as the first liability subject of AI application,and providers and users need to distinguish AI products and services.The uniform attribution ideas lack flexibility,and risk-based approach should be introduced,which not only conforms to the general principle of tort liability attribution,but also implements the concept of hierarchical governance of AI.For high-risk AI,the establishment of no-fault liability can help to fully remedy victims,fairly distribute the risks of AI technology,and avoid the inherent conflict of the attribution system.For limited risk and low-risk AI,presumption of fault liability and fault liability should be applied separately to achieve a dynamic balance between risk and attribution.