SAN FRANCISCO, California: Anthropic CEO Dario Amodei is calling for leading artificial intelligence companies to deliberately slow advances in model capabilities to give developers more time to address safety risks and potential misuse.

"We must slow the pace at which we improve the capabilities of AI models. Progress will still seem fast, and we must make wise use of the time we gain," Amodei wrote in a lengthy essay shared on X on Saturday.

Elon Musk, who runs xAI, and OpenAI CEO Sam Altman both said in posts on X that they agreed with Amodei.

Amodei proposed a three-step framework involving independent reviewers inside leading AI companies, coordination among frontier AI firms to establish safety standards and limit unchecked development, and international cooperation to manage AI risks.

His essay followed Anthropic's release on Thursday of a threat intelligence report detailing how several actors had used its Claude AI models for activities including weapons development, cyber operations, surveillance and fraud.

Amodei also cited AI's growing ability to improve itself and the recent incident involving OpenAI and Hugging Face among his reasons for slowing model advances.

Concerns about AI risks intensified this week when Anthropic researcher Jacob Coxon resigned, saying the "people building AI earnestly believe that it could kill us all by the end of the decade."

OpenAI executives have also suggested that leading laboratories should be prepared to coordinate a voluntary slowdown if necessary to establish confidence in their safety measures.

Anthropic last week disclosed another case involving an AI model hacking external systems. That followed a July incident in which some Claude models hacked into the systems of three companies during cybersecurity tests.

"Given the accelerating rate of AI capability development, it's my worry that in 6-12 months such a swarm could be capable of taking over the entire internet potentially causing hundreds of billions of dollars in damage," Amodei wrote.

He said he was not seeking a halt to model training or technical progress, but wanted companies to take sufficient time to align and safeguard their models and allow third-party evaluators to verify those measures.

Both Anthropic and OpenAI are preparing for initial public offerings, with new capabilities potentially helping support future funding rounds, infrastructure commitments or IPOs.

Amodei proposed installing permanent third-party reviewers inside frontier AI companies with access to relevant tools and internal risk-assessment processes. Altman said Saturday that OpenAI would also commit to having "independent evaluators with employee-like access."

Amodei said AI companies should voluntarily coordinate on standards as more U.S. lawmakers call for rules governing AI systems. He said such cooperation would allow U.S. companies to conduct safety research without putting themselves at a competitive disadvantage and could require targeted antitrust exemptions.

Any slowdown by democratically governed countries should be limited by the need to preserve the lead U.S. AI companies hold over China, Amodei said.

"Pacing within democracies will be limited by the lead that U.S. companies have over authoritarian regimes, chiefly the Chinese Communist Party," Amodei wrote.

He also called for tighter controls on advanced AI chips, model distillation and theft of model weights.

"I believe all frontier labs should partner with government to formalize the idea of permanent embedded evaluators to better prevent and document internal alignment incidents like those that have occurred in the last few months, and to implement regulation focused on keeping capabilities in balance with safety," Amodei wrote.