Anthropic's strategy to slow AI progress gains support from major tech firms

On September 12, 2026, Anthropic CEO Dario Amodei published an essay advocating for a three-step plan to slow AI model development. The proposal, titled 'We Must Pace the Frontier,' received endorsements from OpenAI’s Sam Altman, xAI’s Elon Musk, and Microsoft’s Satya Nadella. Amodei emphasized the need for a balanced approach rather than halting training entirely. His plan includes providing third-party evaluators with permanent access to models, revisiting training foundations, and implementing safety measures. The initiative follows a shift in stance by Amodei, who previously opposed a 2023 pause on AI development. Key factors influencing this change include an independent investigation by METR and insights from Yoshua Bengio, who highlighted misaligned behaviors in AI agents. Nadella stressed the importance of academic involvement and enterprise control over models. The proposal also addresses geopolitical concerns, linking pacing efforts to chip export controls and countermeasures against unauthorized model distillation.