Altman Backs Amodei's AI Slowdown Call, OpenAI to Grant External Auditors Internal Access

Dario Amodei, CEO of Anthropic, published a lengthy article calling for leading AI labs to slow down frontier research and accept external teams on-site to verify safety work.

Anthropic's concrete proposal includes providing external evaluators with desks, badges, company laptops, and the full right to publish their verification results publicly. Amodei also urged the White House to grant a limited antitrust exemption allowing competitors to discuss AI safety without legal risk, arguing the government need only permit such dialogue rather than deeply participate.

Within a day, Sam Altman of OpenAI publicly agreed, stating "I agree with Dario's assessment, we do need to slow the pace of frontier exploration." Altman praised the idea of granting independent evaluators employee-level access and announced OpenAI would implement the same mechanism, with more details coming soon. He revealed that "R&D pace control" had become a core internal topic at OpenAI in recent weeks.

Elon Musk of xAI concurred with a brief endorsement: "Dario is right."

The rapid alignment among top labs marks the first step of Amodei's three-step vision. Notably, neither Amodei's article nor Altman's response mentioned Europe, while an EU Commission technology official emphasized the urgent need for global rules to constrain AI risks.

Brussels has not granted such antitrust permissions since 2004, and no one had previously requested them for AI safety coordination.