In the Era of Artificial intelligence (AI) it is necessary not only to define precisely in the national legislation the extent of protection of personal information and limits of its rational use by other people, to improve data algorithms and to create ethics committee to control risks, but also to establish precise liability (including criminal liability) for violations, related to AI agents. According to existed criminal law of Russia and criminal law of the People’s Republic of China AI crimes can be divided into three types: crimes, which can be regulated with existed criminal laws; crimes, which are regulated inadequately with existed criminal laws; crimes, which cannot be regulated with existed criminal laws. Solution of the problem of criminal liability for AI crimes should depend on capacity of the AI agent to influence on ability of a human to understand public danger of committing action and to guide his activity or omission. If a machine integrates with an individual, but it doesn’t influence on his ability to recognize or to make decisions. In this case an individual is liable to be prosecuted. If a machine influences partially on human ability to recognize or to make decisions. In this case engineers, designers and units of combination should be prosecuted according to principle of relatively strict liability. In case, when AI machine integrates with an individual and controls his abiity to recognize or to make decisions, an individual should be released from criminal prosecution