Both capitals say no within 48 hours
The pacing proposal needed Washington to permit coordination and Beijing to join it. By Monday the White House had dismissed the idea and China's foreign ministry had called it fearmongering.
Gabrielle Richaume via UnsplashA plan to slow the AI frontier had two dependencies outside the companies proposing it. One was an American antitrust waiver permitting rivals to coordinate. The other was Chinese participation. Inside two days, both were refused.
President Trump dismissed the call on Sunday, his first public response. Asked by reporters, he said the United States is leading China in AI and intends to keep it that way, because whoever wins AI wins. He allowed that guardrails were possible, then dismissed the people raising the alarm as negative forces bringing up things that will not happen.
That is a rejection of the premise rather than of the mechanism. The essay Dario Amodei published on Saturday asked government to enable coordination among frontier labs; the answer was that the race is the policy.
Beijing's response arrived on Monday and was colder. Asked about the essay, the foreign ministry spokesperson Guo Jiakun said all parties should work together on AI, and that fearmongering, confrontation and vicious competition will only disrupt global AI governance, which serves nobody's interest.
The irritation is explicable from the text. Amodei called for continued restrictions on sales of advanced chips and chipmaking equipment to China, while also writing that global pacing will require cooperation with China, which he described as the autocratic country with by far the most advanced AI capabilities.
Those two sentences are difficult to hold together. A proposal that asks a country to slow down while arguing it should continue to be denied the hardware is a request for restraint from a position of attempted advantage, and it will be read that way.
The backing the proposal did attract was unusual. Amodei, Sam Altman and Elon Musk all publicly supported reducing the pace over the weekend, which is a rare alignment between three companies that agree on very little else.
It came days after Anthropic published its misuse report covering December 2025 to August 2026, which set out seven categories of disrupted activity including cyber operations, influence operations, surveillance, scams and biological misuse. The alarm was evidenced before it was voiced.
Amodei also made a forecast that will be tested quickly: that a swarm of AI agents could take over the internet within six to twelve months unless researchers agree to slow down. Predictions with dates attached are rare from executives, and this one is checkable by next summer.
China's own position is not inaction. Xi Jinping has called for a global AI governance framework built with the Global South and pitched China as a leader in open-source AI, while the head of its Ministry of State Security has urged accelerating a system for managing AI security risks.
The two leaders are expected to discuss AI governance when they meet later this month, which is now the only forum where the proposal has anywhere to go. A pacing agreement between companies was always going to be a treaty problem wearing a corporate-governance costume.