Abstract: Objective It provides reference for accurate and efficient supervision of medical artificial intelligence industry. Methods By summarizing the main responsibility dilemma of medical artificial intelligence diagnosis products, sorting out relevant researches at home and abroad, the primary responsibility system of medical artificial intelligence diagnosis products is constructed. Results A medical artificial intelligence diagnosis products primary responsibility system with the marketing authorization holder as the "first responsible person" is established, and three safeguard measures are proposed, namely, algorithm transparency and interpretability, classification supervision mode and social co-governance supervision mode. Conclusion The medical artificial intelligence diagnosis products primary responsibility system is helpful to implement the primary responsibility, to build "responsible and beneficial" artificial intelligence, and to realize "self-discipline", "good governance" and "in good order".