Anthropic CEO Proposes Slowing Down the AI Race, Musk and Altman Support

TechFlow reports that on September 13, according to CoinDesk, Anthropic CEO Dario Amodei published an article titled "We Must Control the Frontier" on Saturday, formally calling on the industry to consciously slow the development of frontier artificial intelligence (Frontier AI). Amodei emphasized that the recent pace of model capability advancement has outpaced researchers' understanding and control. While halting technological progress entirely carries risks, pushing forward without constraints is equally reckless.

This viewpoint received rare backing from its competitors. OpenAI CEO Sam Altman responded, "I agree with Dario; we need to control the speed of the frontier," and pledged to follow Amodei's suggestion by introducing independent evaluators within OpenAI who have similar employee access privileges. Elon Musk also briefly replied, "Dario is right," marking a convergence in stance among these three tech giants in the face of AI safety challenges.

Amodei subsequently proposed a phased safety framework: beyond internal independent evaluations, it involves coordinated universal safety standards with participation from democratic governments, with the ultimate goal of pushing for international restriction agreements worldwide targeting dangerous capabilities like "recursive self-improvement." He warned that without necessary safety guardrails, malicious AI swarms with the potential to take over the internet could emerge within months to a year.