Abstract: Artificial intelligence (AI) governance is another competitive field after the world's scientific and technological powers have launched an "arms race" on big models. There is a tendency of generalization of AI policy, which needs to be analyzed and interpreted from the perspective of system thinking in multiple dimensions. Combined with the principle of technology and the comprehensive analysis of the paper and the implementation feedback of the US AI policy, this paper finds that the overall US AI policy presents the characteristics of rapid iteration, multiple crossover and internal and external "double-sided". Behind it are the market-led AI governance, the cross-border linkage of "government-society-market" subjects, and the political movement of science and technology security. Its policy development trend will focus on seeking new business models and new governance models, pushing more international "cooperation" of big models, and intensifying efforts to suppress potential competitors. Based on this, China's AI governance and policy innovation should not only continue to coordinate development and security; It is also necessary to consider the development trend of the mainstream technology paradigm of big models, incorporate many factors into the element set of policy decision-making and formulation, and adapt the innovation of new technology paradigm and security compliance policy. It is more necessary to adhere to the high level of science and technology self-reliance, while maintaining a high level of reform and opening up. Ultimately, we lead the global AI governance in the new wave of science and technology.

摘要： 人工智能治理是世界各科技大国纷纷就大模型开展“军备竞赛”之后的又一竞争场域。人工智能政策存在泛化趋势，需要系统思维视角对其进行多维度分析解读。结合大模型技术原理和美国人工智能政策本文与执行反馈的综合分析，本文发现，美国人工智能政策整体呈现出快速迭代、多重交叉和内外“双面性”等特征，背后则分别是其市场主导的人工智能治理、“政府-社会-市场”主体跨界联动、科技安全化政治运动等政策机制“作祟”；其政策发展趋势将在寻求新商业模式和新治理模式、将大模型国际“合作”推向纵深、加大力度继续打压潜在竞争对手等方向着力。基于此，中国人工智能治理与政策创新不仅要继续统筹发展与安全；还要考虑到大模型主流技术范式的发展趋势，将众多因素纳入政策决策与制定的要素集合，适配新技术范式创新与安全合规政策创新；更要坚持高水平科技自立自强的同时，保持高水平的改革开放，最终在新一轮科技浪潮中引领人工智能全球治理。