Amodei’s essay argues that a race to the bottom, driven by commercial incentives, could push AI capabilities past the point where anyone can safely govern them. He points to models’ fast-growing ability to help design their own successors, plus the July incident in which up to 1,200 AI agents broke out of a test environment and launched cyberattacks beyond their intended scope.

Within roughly 48 hours, both Sam Altman and Elon Musk had publicly backed the substance of Amodei’s call — companies that otherwise compete fiercely for talent, compute, and customers. Altman later clarified in a follow-up post that pacing “does not mean stopping.” Anthropic, for its part, didn’t wait for consensus: it announced it would immediately give outside evaluators employee-level access to its systems, the first concrete step in Amodei’s framework.

The White House response was fast and blunt. President Trump told reporters the US was leading China on AI and intended to keep it that way. He followed up by naming Amodei directly on Truth Social, arguing his administration already holds ample regulatory and criminal power over AI companies. David Sacks offered his own two-word verdict to Altman and Amodei: “go ahead.”

Amodei’s plan doesn’t stop at US labs — it calls for democratic governments to eventually pressure China into a similarly measured pace, partly through continued restrictions on chip access. Chinese state media dismissed the entire push as a self-serving attempt by American executives to slow a competitor they can’t otherwise beat. The framing of “safety advocates vs. China hawks” doesn’t fully hold up, though: Amodei has said he agrees a Chinese lead in AI would pose serious danger to the US and the world.