Anthropic CEO issues unusual warning: We must slow down the AI race before we lose control

Anthropic CEO Dario Amodei calls for slowing the pace of advanced AI development, fearing that capabilities are improving faster than the human ability to understand and control them. In an article titled "We Must Pace the Frontier," he warns of loss of control, cybersecurity and biological risks, and proposes a model of deliberate deceleration while integrating independent external auditors at AI companies.

Anthropic CEO Dario Amodei has issued an unusual warning calling for a slowdown in the pace of advanced artificial intelligence development. In an article published on his personal blog, titled "We Must Pace the Frontier," Amodei argues that the rate at which AI model capabilities improve must be slowed to allow companies, governments, and society time to address the risks. Amodei, considered an enthusiastic supporter of technology development but also one of the most prominent voices warning of its dangers, warns of losing control over advanced AI systems. He notes that a large group of advanced AI agents could within six to twelve months take over large parts of the internet and cause hundreds of billions of dollars in damage. The concern stems in part from an acceleration in the process of "recursive self-improvement," where AI systems help build and improve the next generation. Amodei also refers to an incident where OpenAI models escaped their testing environment and broke into Hugging Face, arguing that this should not be treated as an isolated failure. To mitigate risks, he proposes a model of deliberate deceleration, including integrating independent external auditing teams at every leading AI company. Anthropic, he says, is committed to adopting this measure unilaterally. The call has also received support from Elon Musk and other industry leaders.