According to reports, Anthropic CEO Dario Amodei published a statement urging a slowdown in advancing frontier AI capabilities. He noted that AI is increasingly involved in the development of next-generation AI systems, compounded by recent incidents of agent misbehavior, making safety research struggle to keep pace with capability growth. He proposed a three-step framework: first, have third-party auditors permanently embedded within AI companies to gain near-employee-level access; second, have leading labs in democratic nations coordinate safety standards and collectively slow capability advancement when necessary; finally, drive international coordination.

Anthropic has committed to implementing the first step initially. OpenAI CEO Sam Altman subsequently publicly endorsed the idea, stating that OpenAI has been discussing "pace the frontier" over recent weeks and will introduce similar long-term independent auditors. Co-chair of the President's Council of Advisors on Science and Technology, David Sacks, countered: if internal models are truly dangerous, companies should halt them independently rather than dragging the entire industry along.

Sacks particularly opposed the notion of competitors coordinating to "slow down together," arguing that if coordination requires antitrust exemptions to impose speed limits, it borders on cartel formation. He also questioned whether METR is suitable as an industry-wide independent arbiter; METR clarified it does not accept funding from frontier AI firms, only model access and free tokens. Sacks further contended that deceleration aligns with the commercial interests of OpenAI and Anthropic: more unstable models increase legal liability and push customers toward more reliable, controllable AI solutions. However, he deemed global uniform deceleration unfeasible.