China’s Foreign Ministry rejected Dario Amodei’s proposal to slow frontier AI development after he coupled international safety agreements with measures intended to widen the United States’ lead over China.
“Fearmongering, confrontation and vicious competition will only disrupt the process of global AI governance which serves no one’s interest,” ministry spokesperson Guo Jiakun said at a regular press conference Monday. Guo said countries should work together on AI, responding to a question about Amodei’s essay published Saturday.
Amodei argued that US companies could only slow capability development by an amount smaller than their lead over Chinese rivals. He urged Washington to block sales of advanced AI chips and semiconductor equipment to China, enforce restrictions against chip smuggling and remote access to foreign data centers, stop unauthorized model distillation and strengthen security against model-weight theft.
Those measures could widen the American lead significantly over the next three to five years, Amodei wrote, creating room for US laboratories to spend more time on alignment, monitoring and model evaluations. The prediction is his own. Recent US policy has already moved in several of those directions, including a multi-agency advisory accusing six Chinese laboratories of industrial-scale distillation and a Commerce Department review of remote access to restricted Nvidia chips.
The same essay asked the United States and China to negotiate limits on advanced AI. Amodei described four possible levels of agreement: a ban on AI assistance for biological weapons, shared testing for cyber, biological and alignment risks, a speed limit on AI systems improving their successors, and a full slowdown or pause. He called the first two more attainable and said a full pause was unlikely because either side could evade monitoring.
The state-backed Global Times described the proposal as a “Cold War playbook” intended to restrict China’s AI development through technological barriers and regulatory monopolies. The criticism followed China’s rejection of the US distillation advisory, which its Commerce Ministry characterized as an attempt to suppress competition and monopolize computing power. Beijing said it would take countermeasures if Washington acted against Chinese companies under the advisory.
Amodei’s proposal also calls for a global standards body and compares limits on recursive self-improvement with Cold War arms-control agreements. His framework makes verification a prerequisite for any substantial restraint and begins with outside evaluators embedded inside frontier AI companies. Anthropic has committed to that step, and OpenAI said it would follow.
President Donald Trump separately rejected calls for a US slowdown on Sunday, saying the country should retain its advantage over China. Trump and Chinese President Xi Jinping are expected to discuss AI governance during a meeting planned for September 24.
Sources: Dario Amodei, Associated Press, Reuters
–
By the Control Plane Editorial Team