To pace or not to pace

TL;DR: A new essay from Anthropic CEO Dario Amodei has (for once) brought together AI leaders you’d usually find sniping at each other. The common ground: AI development needs to slow down. But where they—and the White House—disagree is whether government regulation is necessary to make that deceleration stick.

What happened: Amodei’s 3,800-word blog post, which dropped on Saturday, called for a need to “pace the frontier” to avoid developing dangerous, out-of-control AI. The wide-ranging essay has had everyone throwing in their two cents—from rival AI CEOs to the US president and even the Chinese government.

Among the supporters: longtime rival Sam Altman, who welcomed both placing “independent evaluators” inside AI labs (which Anthropic is doing) and a federal AI safety framework. “Every frontier lab must deliver on this, and there is no reason any of us should come to work if we cannot,” he wrote on X. Elon Musk’s comments were more concise (“Dario is right”), and Google DeepMind CEO Demis Hassabis and Microsoft’s Satya Nadella also broadly agreed with the call for a slowdown.

But the Anthropic co-founder specifically wants a referee with teeth: Government regulation that covers all US frontier companies is “the most effective method of pacing,” he wrote, since it would rope in the ones “unwilling to cooperate voluntarily.” Unfortunately for him, there’s a key holdout on this idea: the government.

The regulators against regulation: President Donald Trump’s response to a potential slowdown was clear: “We’re leading China on AI,” he said over the weekend. “And frankly, I want to keep it that way, because whoever wins AI wins.” He then attacked Amodei on Truth Social today and noted that the only guardrail AI needs is “a STRONG AND SMART (High IQ!) PRESIDENT.”

Former White House AI czar David Sacks also had a blunt reply. “Stop pretending you need anyone else’s permission,” he wrote on X, accusing Anthropic and OpenAI of being a duopoly—and that their motivation to slow down wasn’t “purely altruistic.” (Sen. Bernie Sanders, meanwhile, has introduced a new bill banning the development of superintelligent AI—and it includes a 20-year prison sentence for anyone who violates it.)

Please save us from ourselves (and China): If the US pumps the brakes on AI but China doesn’t, that could supposedly hand the country a huge edge. But, per Anthropic, that’s exactly why the industry needs regulation that applies to everyone: to avoid a prisoner’s dilemma where only one lab—or nation—eases off on development while others hit the gas pedal.

And, in Amodei’s view, democratic governments should work together to slow China down (by restricting access to chips and chipmaking equipment) while pressuring it to adopt a more measured pace, too. (An idea Beijing doesn’t exactly seem enthused by, describing these proposed measures as “fearmongering.”)

Bottom line: This might be AI labs’ Oppenheimer moment—the only difference here is they’re not working directly for the government. —WK