Anthropic's CEO wants AI labs to slow down. His rivals agree. Now what?

Key points

• Anthropic CEO Dario Amodei published an essay over the weekend calling on AI companies to slow how quickly they improve their models.
• OpenAI CEO Sam Altman and Elon Musk both voiced public support, a rare moment of agreement among the three rivals.
• Altman said on Saturday that taking OpenAI public this year would be "ill-advised."
• Amodei told CBS News that the plan's biggest weakness is the risk that China keeps developing AI at full speed even if US labs pull back.
• A group of House Democrats called on Speaker Mike Johnson to keep Congress in session until it passes AI safety legislation.

Dario Amodei, CEO of Anthropic, the company behind the Claude family of AI assistants, published a lengthy essay over the weekend laying out a three-step plan for AI labs to slow how quickly they upgrade their models. He argued it could apply the brakes "without sacrificing commercial advantage or the United States' lead in AI."

The essay landed with unusual force because of who endorsed it. Sam Altman, who runs OpenAI, the maker of ChatGPT, and Elon Musk, who leads xAI, both came out in support. These three men run the most prominent AI laboratories in the country and rarely agree on anything. As we reported on 14 September in "AI's biggest bosses agreed to slow down. Almost nobody believes them.", the unity among these chiefs tends to raise as many questions as it answers.

Amodei himself flagged the plan's most obvious vulnerability. Speaking to CBS News, he called it the "toughest dilemma" if China declined to slow its own AI development. A unilateral pullback by American labs could hand Beijing a lead.

Why does this put Anthropic in a difficult position?

Anthropic is reported to be moving toward an IPO, a stock market listing that lets outside investors buy shares. Calling for slower AI development while preparing to pitch your company's growth story to Wall Street is a tension investors will notice.

As CNBC Tech reported, Amodei's deceleration call creates a real balancing act for a company that needs to show investors a path to strong revenue growth while its CEO argues the industry should pump the brakes.

Altman added his own wrinkle on Saturday, saying publicly that an OpenAI IPO this year is "ill-advised." Two of the biggest names in AI, both telling markets to wait.

What does this mean for regulation?

Lawmakers in Washington are paying close attention. A bloc of House Democrats wrote to Speaker Mike Johnson asking him to keep the House in session until it passes concrete AI safety measures. No specific bill text has been voted on yet. Our earlier story "The US Has No Real Plan to Regulate AI, and That Should Worry You", published 16 September, found the federal government's answer to those alarms has so far been to do nothing.

For ordinary people, the policy debate matters more than it might seem. AI safety legislation, if passed, could set rules on which AI products can be sold, how they must label their outputs, and what companies must disclose when something goes wrong. None of that is law yet.

The honest read: the consensus that AI needs guardrails is broader than ever, but agreeing on what those guardrails actually are hasn't started yet. Amodei's essay is a bid to shape that conversation before Congress does it for him.

Common questions

Does Amodei's plan change anything right now?

No concrete change has occurred. It is a published proposal, not a policy or a law, and no government has adopted the framework. No AI lab has formally committed to any specific slowdown.

Should I expect AI products I use to change?

Not in the short term. Any regulatory outcome would take months to years to move through Congress. The tools you use today aren't immediately affected.

Why do AI companies care about China in this debate?

AI models get more capable through intense, expensive training runs. If one country's labs pause and another's don't, the country that keeps going may end up with more powerful tools first. That's the geopolitical risk Amodei flagged.