The dilemma he can't solve

Dario Amodei says the hardest question about slowing AI down is what happens if adversarial nations, China above all, simply keep going.

“I don't know if it's possible, but we should we should try”

A speed limit on progress

Speaking to CBS News' "Sunday Morning," Amodei framed the long-term fix as rival powers agreeing to cap how fast AI improves.

“The more long-term thing would be working together to put a speed limit on the rate of of AI progress”

Why it probably fails

The incentives to pull ahead are enormous, and the military edge that comes with a lead is bigger still. Amodei admits he isn't sure a deal is even achievable.

“the incentives to pull ahead and the military advantage that you get from that are so large”

The essay that started it

Amodei published the proposal on Saturday, urging AI companies to ease off the pace at which they improve their most advanced models.

A researcher walked out

Anthropic's Jacob Coxon quit last week, saying his employer and OpenAI are gambling with people's lives. Another Anthropic researcher put the odds of AI killing all humans above 10%.

“gambling with our lives”

Three steps, no slowdown tax

The plan is designed to avoid sacrificing commercial advantage or the US lead: outside evaluators first, then shared standards among democracies, then talks with authoritarian states.

Anthropic went first

The company has unilaterally adopted step one, giving third-party evaluators employee-level access to verify its safety practices and report incidents.

China's counter-move

Xi Jinping used the BRICS summit in India to call for a consensus-based global AI governance framework and pledged a BRICS open-source AI community.

“consensus-based global AI governance framework”

Two leaders, one meeting

Trump hosts Xi at the White House on Sept. 24, and AI is expected to be on the agenda. Chinese models have been closing the gap on American ones for months.

Rivals line up behind it

Sam Altman, who has feuded with Amodei for years, backed the need to "pace the frontier." Demis Hassabis and Elon Musk endorsed it too.

“Dario's essay points towards the right path forward”

IPOs in the background

Both companies are circling historic listings, but Altman said OpenAI likely won't go public this year, partly over safety concerns.

“Right now would be an ill-advised moment to go public”

Just do it, says the critic

David Sacks argues the two firms already set the frontier and need no one's permission to slow down. Demanding a regulatory framework in exchange, he says, looks like blackmail.

“The easiest way not to build superintelligence is for you to agree not to build it.”