Amodei's AI frontier pacing proposal

Anthropic CEO Dario Amodei published 'We Must Pace the Frontier' on September 12, 2026, proposing three mechanisms to slow AI capability growth: embedded third-party evaluators with near-employee access (which Anthropic is implementing immediately with METR), democratic coordination among frontier labs requiring a government antitrust waiver, and global coordination including chip export controls. Sam Altman of OpenAI, Elon Musk of xAI, and Demis Hassabis of Google DeepMind each endorsed variations of the proposal within hours of publication. Altman suggested OpenAI does not believe it needs to wait for government permission before coordinating with rivals. Legal analyses from Vinson & Elkins and Holland & Knight have since mapped the antitrust landscape: four potential pathways exist for lawful coordination, but Holland & Knight finds existing federal antitrust law was designed to protect market competition, not collective risk mitigation, and ICLE submitted comments arguing antitrust uncertainty is already discouraging joint safety work. A 2025 joint safety evaluation between OpenAI and Anthropic was limited to publicly released models, which ICLE cited as evidence that deeper pre-deployment collaboration is constrained by antitrust concerns. Critics separately questioned whether METR can serve as a genuine check on Anthropic: a critic identified as Argos argued METR lacks meaningful independence, is understaffed, and holds no authority over Anthropic that Anthropic cannot unilaterally revoke. A broader debate over regulatory capture has also emerged, with Zvi and others arguing genuine pacing would slow leading labs more than laggards and is therefore the opposite of capture.