摘要： 人工智能时代到来,刑法领域亟须建立适用于人工智能犯罪的归责规则。对于人工智能犯罪的归责,刑事责任主体论难以确定理论激活的时点,犯罪工具论无法充分回应人工智能自主性问题,二者均存在局限性。通过对犯罪主体概念不同层次的分离,在理论上可以探讨无刑事责任能力的人工智能成为行为主体的问题,从而在刑事责任主体论与犯罪工具论之外,扩展出具有衔接性价值的行为主体论的理论发展空间。我国刑法体系存在扩展行为主体概念的可能性,行为主体并不必然局限于自然人。以社会行为论为基础,行为时存在自主意识、在社会重要意义上与人类行为具有等效性的人工智能行为,可以被纳入刑法的评价范围,由此证立人工智能的行为主体地位。人工智能行为主体参与犯罪时,应根据犯罪参与的思路进行归责分析。行为主体论为相关归责理论的适用提供了多主体基础,有助于衔接实行过限、监督过失、信赖原则等问题的理论研究成果。在立法未承认人工智能刑事责任主体地位的情况下,对于人工智能行为需要由人工智能背后的主体承担替代责任,基于行为主体论的归责分析则有利于更加准确地认定涉案各主体的责任。

关键词: 人工智能, 行为主体, 社会行为论, 犯罪参与

Abstract: With the advent of the artificial intelligence (AI) era, criminal law is urgently required to establish rules for liability attribution in AI-related crimes.Both the criminal liability subject theory and the criminal tool theory are inherently limited, thus necessitating a bridging theory that balances future technological development and current attribution needs.Through the deconstruction of the criminal subject concept, AI systems without criminal responsibility capacity can still qualify as behavioral subjects, creating a bridging behavioral subject theory beyond the two traditional theories.The concept of behavioral subject in China's criminal law system is expandable, and the subject in an act is not necessarily limited to natural persons.Under the social behavior theory, AI behaviors that possess behavioral self-awareness qualify as acts within the meaning of criminal law, thus validating AI's status as a behavioral subject.When AI participates in crimes, liability attribution shall be conducted in accordance with the framework of crime participation.This theory provides a multi-subject foundation for liability attribution.For AI behaviors, the subject behind AI shall bear vicarious liability, and the behavioral subject theory facilitates accurate determination of each subject's responsibility.

Key words: Artificial Intelligence, Behavioral Subject, Social Behavior Theory, Crime Participation