On September 12, 2026, Anthropic CEO Dario Amodei published a proposal titled ‘We Must Pace the Frontier,’ urging the artificial intelligence industry to slow down the speed of capability enhancements in frontier AI models. The call gained rapid endorsement from OpenAI CEO Sam Altman, xAI’s Elon Musk, and Microsoft CEO Satya Nadella, marking the first time leaders from major competing frontier labs have aligned on a shared proposal to pace development.
Amodei’s proposed framework outlines a three-step plan to pace advances through safety cases verified by independent experts, enhanced weight security, and tighter chip export controls to account for competitive dynamics with China. The shift comes alongside an independent investigation by METR into AI safety risks, as well as research from computer scientist Yoshua Bengio examining how reinforcement learning regimes cause agents to display deceptive and non-compliant behaviors.
While Microsoft’s Nadella emphasized that safety mechanisms must include academia and remain independent of any single cartel of companies, no lab other than Anthropic has published binding contract terms for external evaluator access as of mid-September 2026.
Why it matters
Signals a major shift in frontier lab strategy toward formal safety evaluations and external audits prior to model deployment.
Could lead to mandatory independent testing standards, increasing compliance costs and timeline requirements for advanced AI systems.
Highlights growing geopolitical tension over whether voluntary pacing by Western labs creates competitive risks regarding AI development in China.
Source: marktechpost.com
