Anthropic chief executive Dario Amodei has urged democracies to slow the frontier of artificial intelligence and to seek cooperation with China, in a blog post titled "We Must Pace the Frontier".
The call carries a catch.
Amodei tied any global pacing to steps that would preserve a lead for the US and its allies, including tighter controls on chip smuggling and limits on unauthorised model distillation, a technique used to copy a rival model's capabilities.
Beijing was unimpressed.
Mongering the fear
"This is fear-mongering, confrontation, and vicious competition," said Guo Jiakun, a spokesperson for China's Ministry of Foreign Affairs.
Chinese officials and researchers said they share the underlying safety concerns.
But they view the proposal as a bad deal, one that treats China as a rival to be constrained rather than an equal partner.
Amodei himself acknowledged he did not know how to negotiate with Beijing, a candid aside that underscored the diplomatic challenge.
The tension feeds a wider Silicon Valley debate that some outlets have labelled "AI psychosis", describing engineers and executives swinging between panic-driven safety warnings and competitive urgency.
There is a near-term test.
Early talks
Preliminary talks are reportedly underway ahead of Chinese leader Xi Jinping's planned visit to Washington later this month.
That will show whether the two sides can turn shared technical worries into verifiable commitments.
