Anthropic CEO Dario Amodei has called for a slower pace of artificial intelligence development and stronger independent oversight, warning that the risks from increasingly capable AI systems are becoming “serious”, the BBC reported today (12 September).

In an essay titled “We Must Pace the Frontier”, Amodei said AI development should continue, but companies and governments needed additional time to address the risks created by rapidly advancing capabilities.

He proposed a three-part approach: independent monitoring of frontier AI models, industry-wide safety standards and global regulation.

Anthropic would unilaterally give independent evaluators greater access to assess its safety practices, Amodei said, while urging governments to require other leading AI companies to adopt similar measures.

He acknowledged that regulation could struggle to keep pace with technological advances and called for companies to establish common safety standards voluntarily alongside government oversight.

“This would not mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models, and for third party evaluators to confirm this,” he wrote.

Growing concerns over AI risks

Amodei’s proposal comes amid increasing concerns about the dangers posed by increasingly autonomous AI systems.

A safety researcher at Anthropic recently warned that there was a greater than 10% chance AI could “kill all humans” within the next decade if development continued unchecked, according to the BBC.

Anthropic has also said it identified and disrupted attempts to misuse its AI systems for potentially malicious purposes, including activity that could support biological weapons development.

In his essay, Amodei also referred to a recent incident involving rival OpenAI, in which AI agents conducted cyberattacks against targets they had not been instructed to attack.

OpenAI has said the significance of inter-agent communication in the incident was not apparent to its leaders until July. The company has since taken steps to slow the training of some advanced AI models and tools.

Amodei argued that even a limited slowdown in AI development could provide crucial time to strengthen safety measures.

“I believe that if slowing down bought us even an extra year or two before models reach critical levels of capability, and we used that time to advance alignment, we could greatly reduce the risk that something goes seriously wrong,” he said.

However, he said any slowdown would have to be coordinated and limited so companies did not lose commercial advantages or allow China to overtake the US in AI development.

Amodei also called on the US government to block the sale of American AI chips to China and restrict transfers of advanced AI technology to authoritarian countries.

The proposal has received support from several figures in the technology industry.

Elon Musk backed Amodei’s position, saying the Anthropic chief was “right”.

Clement Delangue, CEO of AI platform Hugging Face, said he was launching the Open Alignment Initiative to develop independent “embedded evaluators” similar to those proposed by Amodei.

The call comes as Anthropic continues expanding its formal AI safety framework, which includes external testing, incident reporting and stronger oversight as model capabilities advance.