An essay by Anthropic CEO Dario Amodei calling for slower AI development “may look like a rational statement focused on safety risks”, but it is really a “Cold War playbook” targeting China — so argues the state-backed Global Times in an editorial reported by Reuters on September 14. Anthropic’s essay frames the same proposal the opposite way: a framework to pace frontier model capabilities and “create more time to manage mounting safety risks”.
Each side is reading the other’s motive, and the gap between those readings is now the story.
What the Global Times actually said
The editorial states that the essay’s true agenda is “to attempt to curb China’s AI development through technological barriers and regulatory monopolies, uphold Washington’s monopolistic hegemony in cutting-edge technology, and exclude China from the global AI governance system”. It calls this “silent AI Cold War” “hypocritical and short-sighted”, and argues that excluding China from global AI innovation would “significantly increase the trial-and-error costs and risks of loss of control in global AI development”.
That last point is a safety argument of its own — Beijing’s state media is claiming that a gated, US-centred frontier would be more dangerous, not less, because the excluded labs would develop anyway and the coordinating centre would have less visibility. It is a mirror image of Amodei’s own logic, pointed in the opposite direction.
Why the essay drew fire where the warnings did not
The Global Times editorial did not focus its fire on the risk claims that have driven this week’s news cycle — the warnings from Anthropic researchers about catastrophic outcomes, which US lawmakers have cited while urging new rules for AI systems — the same warnings behind the 1,100 lab staffers’ open letter asking Washington to back a slowdown. What it targeted was the essay’s policy apparatus. Amodei’s piece urged the United States to strengthen chip export controls on China and to crack down on what Anthropic has alleged is model distillation by Chinese AI labs — allegations Chinese officials and state media have rejected before, and which we covered when Beijing dismissed the distillation claims.
Amodei himself acknowledged the friction on Sunday. Speaking to CBS News, he called the “toughest dilemma” of his proposal the possibility that China and other adversarial nations simply choose not to slow down — a scenario that would hand a pacing lab’s rivals an open road. From Beijing’s vantage point, the essay concedes its own leverage problem: a slowdown only binds those who agree to it, so the essay leans on Washington’s export-control regime to make restraint compulsory for everyone else. That is the passage the Global Times read as a “playbook”, and the essay’s export-control language is what it points to.
The dispute is now scheduled diplomacy
The exchange lands two weeks before it can matter procedurally. The United States and China are planning to discuss frontier AI safety risks in mid-September, according to sources cited by Reuters, and the issue could be raised during a summit between Presidents Donald Trump and Xi Jinping on September 24. Washington’s own position cuts against the labs: Trump rejected the slowdown calls on Sunday, saying “very negative forces” were raising exaggerated concerns and that the United States is leading China in AI — so the US delegation would arrive backing acceleration while Chinese state media accuses it of demanding restraint.
For Anthropic, the Global Times rebuke is the first state-level rejection of the pacing agenda — and a test of whether a safety argument can survive being read as a trade policy. For the labs’ slowdown consensus, the episode shows the harder problem Amodei named on Sunday: pacing is only stable if everyone paces, and at least one superpower has now said publicly it won’t be told to.
