Abstract：The Third Plenary Session of the 20th Central Committee of the Communist Party of China proposed to establish an artificial intelligence safety supervision system,and it is urgent to strengthen the theoretical exploration of the relevant supervision system.Through a systematic study of the safety supervision system of generative artificial intelligence in China,the European Union and the United States,it is found that China's supervision system has deficiencies in subject,norms,tools and accountability.The EU builds an integrated regulatory system with internal co-ordination,security and credibility,risk orientation and penalties for violations as the core.Based on the political system,rights protection,development orientation and market incentives,the United States has formed a distributed regulatory system.In the future,China should strengthen the coordination of the main body,establish a multi-connected interaction mechanism,and enhance the supply of regulatory resources.China should optimize the regulatory system,based on horizontal classification and vertical stratification path,and gradually improve and unify regulatory norms.China should expand tool configuration,strengthen the construction and application of quality tracking evaluation and multi-level risk monitoring tools.It advocates prudent accountability,optimizes the mechanism of damage imputation and liability reduction,realizes the rational allocation of responsibility and builds an innovative fault-tolerant space.
何野. 生成式人工智能安全监管制度优化:欧美经验及其启示[J]. 中国科技论坛, 2025(10): 180-188.
He Ye. Optimization of Safety Supervision System of Generative Artificial Intelligence: European and American Experience and its Enlightenment. , 2025(10): 180-188.
2025, 
