Deep Dive: Who Should the U.S. Engage With in China on AI Policy?
Key Takeaways
- ▸China's AI governance is fragmented across multiple agencies with competing interests; the U.S. must strategically choose counterparts based on specific policy topics
- ▸The CAC is the logical lead for technical standardization and mandatory testing of frontier models; the NDRC should coordinate broader AI policy discussions
- ▸Bilateral dialogue should focus on communicating about emerging risks and sharing safety-enhancing practices (testing/evaluation approaches) rather than formal enforcement agreements
Summary
Following the May 2026 Trump-Xi summit, the U.S. and China announced plans for government-to-government dialogue on artificial intelligence. Matt Sheehan's comprehensive analysis maps China's complex AI bureaucracy and identifies the optimal Chinese interlocutors for different aspects of AI governance. The piece argues that the Cyberspace Administration of China (CAC) is best suited for conversations on mandatory testing and evaluation of frontier AI models, while the National Development and Reform Commission (NDRC) should lead on broader AI policy coordination and governance strategy. Sheehan emphasizes that productive dialogue is possible if focused on shared threats and mutual testing/evaluation best practices, though domestic AI governance actions within each country will likely remain more consequential than bilateral agreements.
- Domestic AI governance actions within the U.S. and China will remain more impactful than international bilateral accords in the near term
