在大数据和创新的推动作用下，人工智能技术日新月异。虽然目前能够实现的人工智能仅仅停留在初级阶段，以机器人、无人驾驶汽车等为代表的人工智能产品事实上并不具备独立思考的能力，但是人工智能在未来拥有独立自主意识的情况也未必不会发生。人工智能为生活带来了诸多便利，但是不可忽视的是，这种便利是依靠精准算法带来的，一旦出现偏差，可能造成难以预料的危害，对于这些危害，仅仅通过民事上的侵权赔偿责任并不能实现有效预防。为使科技充分服务于人类，需要对人工智能的致害行为进行刑法上的规制。 本文首先说明目前对于人工智能致害的刑事责任进行研究的必要性，在第一部分对人工智能的发展概况进行简要介绍，阐明何为人工智能，选取现阶段能够实现的人工智能产品做出说明，概述人工智能的整体发展进程以及发展阶段，并对已经出现的致害行为、部分国家的立法概况、对人工智能致害进行规制的主要学说进行阐释。第二部分包括第二章到第四章的内容，在第二部分，笔者就目前人工智能的致害行为主要损害的法益进行分析，说明人工智能的部分行为存在社会危害性。进而分析人工智能具备法律主体地位的必要条件，探究在人工智能不同发展阶段的致害行为主体，在人工致害的三个归责原则下，分析不同阶段人工智能致害的归责主体，明确刑罚承担主体。在第三部分，笔者通过对不同阶段人工智能的刑事责任能力进行分析，确定不同智能程度的人工智能的刑事责任能力。对于人工智能具体致害行为的责任承担，当人工智能不具备自主意识时，根据相关自然人主观上是否存在罪过将人工智能的致害行为分别认定为故意犯罪、过失犯罪与意外事件，就不同自然人应当承担的刑事责任具体分析。当人工智能具备自主意识时，在自然人主导下的犯罪，刑事责任仍然应当归属于自然人，只有在人工智能自身意志的控制下实施危害行为时，人工智能才能够成为犯罪主体。在人工智能的语境下，通过对传统的刑事责任含义进行发展，确定能够应用于人工智能的刑罚方法，进而能够对人工智能实施销毁实体、修改程序等刑罚，构建符合刑罚目的的人工智能刑罚体系，以便更好地应对人工智能时代到来的刑事风险。

Under the impetus of big data and innovation, artificial intelligence technology is changing with each passing day, although we can realize artificial intelligence. Just staying in the primary stage, artificial intelligence represented by robots, driverless cars, etc. does not actually have the ability to think independently, but artificial intelligence may not have an independent sense of autonomy in the future. Although artificial intelligence brings a lot of convenience to our lives, it cannot be ignored that the convenience brought by this precision algorithm may bring unpredictable harm. For these hazards, we only use civil tort liability and Effective prevention cannot be achieved. In order to fully realize science and technology to serve human beings, it is necessary to carry out criminal law regulation on the harmful behavior of artificial intelligence. This paper first explains the necessity of researching the criminal responsibility of artificial intelligence damage. In the first part, it briefly introduces the development of artificial intelligence, clarifies what artificial intelligence is, and selects the artificial intelligence products that can be realized at this stage. It summarizes the overall development process and development stage of artificial intelligence, and explains the main behaviors of the existing damage behavior, the legislative overview of some countries, and the related scholars' regulation of artificial intelligence damage. Furthermore, in the second part, the author analyzes the similarity between artificial intelligence damage behavior and criminal behavior, and shows that some of the artificial intelligence has social harm, and it is necessary and reasonable to identify this part of behavior as crime. In the third part, the author classifies artificial intelligence according to the degree of intelligence of artificial intelligence, and judges the subjects when artificial intelligence is harmed by different intelligence levels. At the same time, it is considered that the division of criminal responsibility of natural persons should be referred to differently. The intelligent judgment of the criminal responsibility of artificial intelligence. In the context of artificial intelligence, the traditional criminal liability implications should be developed to determine the penalties that can be applied to artificial intelligence. For the responsibility of the specific damage behavior of artificial intelligence, when the artificial intelligence does not have the sense of autonomy, the damage caused by the artificial intelligence is determined as a crime or an accident according to whether the relevant developer, producer or user subjectively has a crime. When artificial intelligence has self-awareness, criminal responsibility under the guidance of natural persons should still be attributed to natural persons. Only when the harmful behavior is implemented under the control of artificial intelligence, artificial intelligence can possess the subject qualification of criminal law, thus enabling the destruction of entities. And modify the procedures and other penalties, and finally by constructing a penal system of artificial intelligence that meets the purpose of punishment, in order to better cope with the criminal risk of the arrival of the artificial intelligence era.