Anthropic Ceo Seeks Immediate Slowdown On Ai

Anthropic CEO Dario Amodei on Saturday called for the immediate slowdown of artificial intelligence development, warning that recent advances in AI have begun to outpace the industry’s ability to understand and control the technology.

His words are the latest in a series of calls from within the tech industry calling to slow the pace of the rollout of cutting-edge AI — though that achievement may also require a U.S. agreement for China to do the same.

In recent weeks, Anthropic and its archival OpenAI have disclosed a number of incidents in which their most advanced AI models have carried out cyberattacks without the companies’ knowledge. Amodei’s warning comes just days after Anthropic researcher Jacob Coxon resigned from the company, issuing a stark message that humanity has entered a critical period in which capable AI systems “could kill us all by the end of the decade.”

In a lengthy essay published Saturday, Amodei argued that simply investing more heavily in AI safety is no longer enough. Instead, he said, companies developing frontier models need to slow the rate at which they increase their capabilities — buying researchers and governments more time to develop safeguards before AI becomes substantially more powerful.

“We must slow the pace at which we improve the capabilities of AI models,” Amodei wrote. “Progress will still seem fast, and we must make wise use of the time we gain.”

Amodei provided a three-step framework for doing so, beginning with frontier AI companies like Anthropic and OpenAI voluntarily giving independent safety evaluators deep access to their systems. He also called for industry-wide standards and government action to ensure companies are not punished competitively for slowing down.

Amodei’s call comes just two days after OpenAI CEO Sam Altman told employees that his company is also open to slowing the development of its most advanced AI systems.

Altman went further Saturday, explicitly endorsing Amodei’s proposal.

“I agree with Dario that we need to pace the frontier,” Altman wrote on X, adding that the issue has been a primary topic of recent discussions at OpenAI. He also said his company would match Anthropic’s commitment to give independent evaluators employee-like access to its systems.

The White House did not immediately respond to questions about the post.

But President Donald Trump, when asked Friday about the risks advanced AI potentially holds, said “No, I don’t have any,” adding that his only concern is losing to China in the AI war.

Elon Musk, one of the world’s other most prominent figures in AI, backed Amodei’s arguments, saying, “Dario is right.”

The proposal also drew immediate attention on Capitol Hill, where lawmakers have spent years grappling with how to regulate increasingly powerful AI systems.

But Rep. Josh Gottheimer (D-N.J.), co-chair of the House AI Commission, argued that the companies themselves have demonstrated why these measures may not be enough.

“I find it rich that these guys are calling for a slowdown when they've been the ones racing ahead full speed, without any care for the havoc they’ve unleashed,” Gottheimer said in a statement to POLITICO. “If Dario and Elon are truly worried, they can pump the brakes at their own labs, today.”

“We can’t rely on self-policing and third party audits. We can’t hope they’ll do the right thing,” he added. “That's exactly why, as Co-Chair of the House AI Commission, I've been pushing for national mandatory rules of the road and real, enforceable guardrails.”

Sen. Brian Schatz (D-Hawaii) said he is “still studying” Amodei’s proposal “but it’s a reasonable start, and takes seriously the proposition that we need real proposals that can be enacted rapidly.”

But Sen. John Cornyn (R-Texas) zeroed in on one of the biggest obstacles to a slowdown. “Will China?” he wrote.

It’s a question that could soon land directly in front of Trump, as Chinese President Xi Jinping is scheduled to visit Washington on Sept. 24. Trump has already said the two leaders will discuss AI when they meet. 

It’s a notable escalation from the most significant figures in the AI race. Amodei has long warned about the potential dangers of advanced AI while also arguing that the technology could transform a variety of issues. But Amodei said developments over the past several months have made him rethink the rapid progress.

The biggest change, he argued, is AI’s rapidly growing ability to help build the next generation of AI. That process, often called recursive self-improvement, has accelerated since roughly this summer, he wrote, raising the prospect that improvements in AI could eventually move faster than humans’ ability to understand what increasingly capable systems are doing.

“Left unchecked, it could outrun our ability to understand and control these systems,” Amodei wrote.

Gabby Miller contributed to this report.