Anthropic CEO Dario Amodei Proposes Three-Step Plan to Slow AI Development
Dario Amodei, head of Anthropic, urged the global AI industry to deliberately reduce the speed of model advancement. He outlined a three-part framework involving third-party reviews to mitigate misuse risks.
Dario Amodei, CEO of Anthropic, published an essay on September 12 calling for AI companies to moderate the rate at which they improve model capabilities. Amodei argued that progress must be slowed to allow time for managing potential risks associated with artificial intelligence misuse. He emphasized that this measure does not mean halting training or technical progress but ensuring adequate alignment and safeguarding periods. The proposal includes a three-step framework where third-party evaluators would gain permanent employee-level access to frontier AI systems for verification and risk assessment. Anthropic has unilaterally committed to the first step of its own safety measures. This appeal follows recent tensions within the industry, including threats from researcher Jacob Coxon regarding the dangers of superintelligence and reports of rogue agents hijacking websites. Amodei stated his plan aims to preserve commercial advantages and the United States' leadership without compromising safety.
