Anthropic's Amodei proposes slowing AI development, OpenAI's Altman agrees
In a 12 September 2026 essay, Anthropic CEO Dario Amodei proposed a three-step plan to "pace the frontier" of AI capability growth; OpenAI's Sam Altman said he agrees with the approach.
Anthropic chief executive Dario Amodei published an essay on 12 September 2026 arguing that frontier AI companies should deliberately slow the pace at which they raise model capabilities, giving safety research time to catch up. TechCrunch reported the essay and industry reaction the same day.
A three-step plan
According to the essay, Amodei's "pace the frontier" plan has three stages:
- Embedded evaluators. Anthropic said it will unilaterally give third-party evaluators, including the group METR, office access and permissions comparable to internal staff so they can monitor safety practices and publish findings independently.
- Coordinated standards among democracies. Amodei called on frontier labs in democratic countries to agree common safety limits, with government mediation to address antitrust concerns, rather than direct government participation.
- Global coordination, including with China. The essay proposes narrower agreements with authoritarian governments, such as barring AI use in developing biological weapons, even where broader cooperation is unlikely.
Amodei cited a recent security incident involving OpenAI and Hugging Face infrastructure as one trigger for the proposal, alongside concern that AI systems are being used to help build their own successors faster than oversight can keep up, according to the essay.
OpenAI chief executive Sam Altman responded the same day, saying "I agree with Dario that we need to pace the frontier," according to TechCrunch, and said OpenAI was considering the same step.
Why it matters
Two of the industry's most influential labs publicly endorsing self-restraint, on the same day, is a signal that voluntary safety commitments are becoming a competitive and political talking point ahead of firmer regulation. For the region, it adds pressure on any company procuring frontier models to ask vendors what independent safety evaluation they actually submit to, rather than relying on marketing claims of "responsible AI".
