Centralization, Risk, and Innovation: A Two-Axis Framework for China’s Generative AI Governance in Comparative Perspective

This article analyses China’s emerging framework for the governance of generative artificial intelligence (AI), placing it in comparative perspective with the United States (U.S.) and the European Union (EU). Drawing on original two-year data of 801 algorithm and generative AI filings (August 2023 to August 2025), it demonstrates how China has developed a centralized, state-first model, built on mandatory national filings supplemented by derivative local registrations. This arrangement enables the state to mobilize resources, counter local protectionism, and endorse national champions such as Baidu, Tencent and DeepSeek. Yet the expansive definition of “high-risk” systems—anchored in “public opinion attributes” and “social mobilization capabilities”—produces an absence of effective risk differentiation, subjecting nearly all public-facing AI services to stringent filing and content moderation. By contrast, the U.S. relies on a decentralized, innovation-first model rooted in federalism and market dynamism, while the EU adopts a hybrid but risk-first approach under the 2024 AI Act. The comparative analysis shows that China’s centralized system has proved effective in the short term for industrial policy and global competitiveness, but risks long-term constraints on innovation. The article reconceptualizes AI governance as a two-axis design space—centralization versus decentralization, and innovation promotion versus risk prevention—and argues that proportionate, transparent, and differentiated regulation is essential for sustainable governance of generative AI.