Abstract: The affirmation of AI criminal responsibility subject hides the problems of floating the problem， simplifying the strategy， and neglecting the legal principle. The current negative theory of AI criminal responsibility subject has the problem of neglecting the specific context of affirmation and the lack of critical power.It should be legalized and philosophically proved. Neither the weak AI products nor the strong AI products should be the subject of criminal responsibility. They are only intelligent mechanical products that humans can control and that rely on electric energy for survival. They should be used as tools for human beings to achieve free and comprehensive development. Man is the measure of all things.The affirmation theory insists on a mechanical “human”. The core part of the application of criminal law and the basic principles of criminal imputation will not change fundamentally due to the emergence of AI technology. The risk of the development of AI technology depends not only on the top-level design of the algorithm， but also on the integration of rule of law management into every aspect of AI development to prevent negative consequences. Legal persons must acknowledge the limitations of their knowledge， maintain moderate restraint， and must not mistake emotion or association for “reality”.

Key words: the affirmation of AI criminal responsibility subject, the negative theory of AI criminal responsibility subject, jurisprudence proof, philosophical proof