China presses its case to lead AI safety rules

At the World Artificial Intelligence Conference in Shanghai, China presented itself as a new center of gravity for AI governance while the U.S. under President Donald Trump has shifted toward AI dominance and lighter regulation. Xi Jinping announced a new Shanghai-based governance body comprising 29 countries and promoted open-source AI as a more accessible alternative to proprietary U.S. systems.

The repositioning follows a sharp change in Washington. During Joe Biden’s presidency, U.S. officials tried to steer talks with China toward frontier risks, but diplomacy stalled. Under Trump, established federal guardrails were dropped, the Paris AI safety summit became the “AI Action Summit,” and Vice-President J.D. Vance warned against “excessive regulation.” Western safety advocates involved in Track II dialogues say Chinese regulators have become more receptive to terms and frameworks common in global AI safety debates.

China’s approach still centers largely on near-term harms such as deepfakes, algorithmic bias, children’s use, and AI companions, while its system remains less developed on catastrophic frontier risks. Powerful model releases and internal hacking tests by U.S. firms have pushed Washington back toward security concerns, including a voluntary review process announced in June. The U.S. and China are set to restart discussions on AI safety and governance in September, but competing visions for global standards leave any agreement uncertain.