A Nature analysis notes: China issued as many national-level AI regulations in the first half of 2025 as it did in the previous three years combined.
Regulatory framework
Since 2022, China has built a fairly comprehensive AI regulatory system:
- Generative AI models must submit to regulatory review before launch
- AI-generated content must carry visible, non-removable watermarks
- Developers bear responsibility for model outputs
In August 2025, the State Council released the AI Plus Action Plan, targeting:
- 70% AI penetration in key industries by 2027
- 90% by 2030
Covering six areas: scientific research, industrial applications, consumer services, public welfare, security governance, and international cooperation.
Standards implementation
The Measures for Labeling AI-Generated Content, effective September 1, along with the mandatory national standard GB45438-2025, are among the world's earliest AI content labeling regulations.
Global comparison
Three main regulatory approaches currently exist:
- EU: Rights- and risk-based tiered regulation (AI Act)
- US: Preference for voluntary standards and industry self-regulation
- China: Inclusive cooperation plus state sovereign control over data and AI deployment
Shifting competitive landscape
A telling data point: the global market share of Chinese LLMs surged from 3% to 13% in just two months, driven largely by DeepSeek.
The path Chinese AI companies are taking is becoming clear — rapid iteration within the regulatory framework. Regulation is not a stumbling block but rather a screening mechanism: companies that can make their business models work within compliance boundaries gain a compliance advantage when going global.
Nature's advice is that other countries need to "actively engage" rather than "avoid" China's AI governance process. Regardless of one's stance, China's speed and execution in AI regulation have indeed taken the lead.
Sources: CocoLoop, Nature analysis article