Sam Altman agrees AI needs to slow down

OpenAI and Anthropic are rarely on the same page, but slowing down frontier AI seems to have brought the rivals together.

OpenAI CEO Sam Altman publicly backed Anthropic CEO Dario Amodei's call to pace frontier AI development. Altman revealed that pacing has been a primary topic of internal discussions at OpenAI over recent weeks.

Handing outsiders employee badges

Amodei recently published an essay arguing the industry needs to slow down, unilaterally pledging that Anthropic will grant third-party evaluators permanent, employee-level access to its systems. The goal is to verify safety compliance, report incidents, and evaluate model alignment right during training.

Altman called the concept a great idea and confirmed OpenAI will follow suit, adding that more details will be shared soon.

Why it matters

Leading AI labs have spent years locked in an unrelenting sprint where ship dates trumped almost everything else. Having both OpenAI and Anthropic openly discuss pacing the frontier marks a genuine shift in tone across the industry.

We think granting outside auditors internal, employee-grade access during training is a much smarter approach than basic post-launch red teaming. Still, real accountability hinges entirely on the fine print. It remains to be seen who these evaluators actually are, how independent they stay, and whether this voluntary slowdown holds once either side hits their next major breakthrough.