Altman and Musk Back Amodei's AI Pacing Call Within Hours

Dario Amodei published an essay on September 12 arguing AI development needs to slow down. By the end of the day, Sam Altman and Elon Musk had both agreed.

That's a fast yes from three labs that spend most of their time racing each other on benchmarks and releases. Independent evaluators with staff-like access would put outsiders inside Anthropic, OpenAI and Google DeepMind, watching how these systems get built before release.

Altman said committing to independent evaluators with employee-like access "is a great idea," and that OpenAI would follow. Musk posted that "Dario is right." Hassabis backed the direction too, pointing to a separate DeepMind proposal for an industry-wide frontier AI standards body, per Politico's report.

Antitrust commentators flagged the obvious problem. Three companies that dominate frontier AI are now agreeing on shared rules for frontier AI, raising the bar for anyone smaller trying to catch up. Shared rules cost the leaders less than they cost a challenger trying to close the gap.

Politico's report doesn't say what "employee-like access" means in practice, whether these evaluators can block a release or just observe one. No Chinese lab has signed on to anything, so any pacing agreement here covers only part of the industry. A separate open letter pushes a different fix. It would skip the evaluators and require any model sold to the public to open its weights.