Anthropic CEO Dario Amodei warns AI is advancing faster than expected

Anthropic CEO Dario Amodei is sounding less like a promoter than a man trying to put a hand on the brake. In a CBS News interview, he said the pace of AI progress has surprised him and that the industry spent too long pretending the risks weren’t real.

His warning wasn’t abstract. Amodei said AI has been moving along an exponential curve that is now getting steeper, and he called that a sign to slow down. Not shut it all off. Not panic. Just stop treating every new model like a victory lap.

That message came with a concrete proposal. In an essay published Saturday, Amodei argued that the whole tech industry should work together to slow the pace of AI capability growth. He also said Anthropic will give independent evaluators permanent, employee-like access to its models so third parties can check whether the company is actually following the safety practices it says it will follow.

Amodei doesn’t think that’s enough on its own, and he said so plainly. He supports federal regulation, including ideas now floating around from a ban on superintelligence to a mandatory kill switch. He said a total ban would block major benefits, including possible medical cures, and argued that the technology would still be built elsewhere, including in authoritarian countries.

The hardest part, he said, is the international race. He pointed to nuclear arms agreements with the former Soviet Union as a model, though he admitted the incentives to race ahead are enormous and that he doesn’t know whether a global speed limit on AI progress is even possible. Still, he said the effort should be made, because if AI is built the wrong way, the odds of something bad happening rise sharply.