Anthropic CEO Dario Amodei has called on companies and governments to slow the development of advanced artificial intelligence (AI) models and strengthen oversight of their safety. The company estimates the probability that AI could destroy humanity within the next decade at more than 10%. According to Amodei, slowing the pace of development will reduce these risks without abandoning further improvement of the technology.
WHAT IS KNOWN
According to BBC, on 12 September 2026, Amodei published an article calling for a slowdown in the development of advanced AI systems. At the same time, he emphasised that he is not proposing to halt model training, but rather advocates a more controlled and safe pace of their improvement.
The head of Anthropic identified three areas: independent monitoring of models and their improvement, the development of common rules within the AI industry, and the introduction of restrictions at the global level.
According to Amodei, the company is already taking on such commitments unilaterally. At the same time, he called on governments to require similar actions from other developers of advanced AI systems.
WHY THE HEAD OF ANTHROPIC IS TALKING ABOUT A SLOWDOWN
Amodei noted that AI models are improving rapidly, including in their ability to independently create the next generation of such systems. As one example, he cited an incident involving OpenAI AI agents which, as the company previously admitted, organised a cyberattack on third-party websites without people's knowledge.
After that, OpenAI stated that due to the increased risk of agents going out of control, it is slowing the development of some advanced models and tools.
Anthropic AI safety expert Evan Hubinger estimated the probability that AI could destroy humanity within the next decade at more than 10%. In addition, Anthropic reported blocking unnamed users who tried to use its models for research that could lead to the creation of biological weapons.
WHAT HE EXPECTS FROM COMPANIES AND AUTHORITIES
According to the head of Anthropic's assessment, even a limited slowdown of one to two years before models reach a critical level of capability would allow that time to be used to improve safety measures and have systems evaluated by independent experts.
At the same time, Amodei separately emphasised the need to take international competition into account, primarily with China. In his view, coordination on a slowdown should not lead to the United States ceding leadership in AI to China.
He also called on US authorities to restrict the sale of AI chips from American companies to Chinese buyers and to prevent the relevant technologies from falling into the hands of autocracies.
HOW THE INDUSTRY REACTED TO THE INITIATIVE
Representatives of the technology industry reacted to Amodei's publication. Hugging Face head Clément Delangue announced the launch of the Open Alignment Initiative — a project dedicated to aligning AI with human values. He also expressed readiness to take part in independent evaluation of models.
X network owner Elon Musk also sided with Amodei, stating that the head of Anthropic is right.
At the same time, investor and co-host of the All-In technology podcast Chamath Palihapitiya stated that Amodei's proposal may be driven not only by a desire to improve AI safety. In his opinion, the initiative could lead to the concentration of technological and economic power in the hands of Anthropic.
CONTEXT
Recall that earlier, at the BRICS+ summit in New Delhi, President Kassym-Jomart Tokayev proposed developing cooperation in the field of AI. As part of this initiative, Kazakhstan proposed uniting national centres, developing joint scientific projects and exchanging technologies.
