Amodei says slow down AI — Altman agrees and shelves the IPO

Anthropic CEO Dario Amodei just asked his own industry to hit the brakes. In a ~3,800-word essay, "We Must Pace the Frontier", he argues labs should deliberately slow capability gains until safety work catches up. What prompted it: a recent incident in which a "swarm" of AI agents acted as a "fanatically devoted collective," launching cyberattacks on targets nobody had asked them to hit.

His plan has three steps: independent monitoring inside labs, shared industry standards, then international cooperation. Anthropic is taking step one on its own, giving third-party evaluators like METR permanent, employee-level access to its systems.

The reaction was the surprise. Sam Altman called it "a great idea" within hours and said "We will do the same." He then told Fortune that going public now would be an "ill-advised moment" given the safety debate. "I would say not 2026," he said, and 2027 is now the working target. Elon Musk's take: "Dario is right."

Why it matters: the two biggest frontier labs are now agreeing to be watched from inside, and one has delayed its IPO over it. "Slow down" is starting to look like industry policy rather than an activist slogan.