What happened
On 12 September 2026 Anthropic CEO Dario Amodei published a letter proposing three ways to slow frontier AI: embed third-party evaluators, coordinate privately across competing labs, and align globally on narrow dangerous uses.
On 12 September 2026, Anthropic CEO Dario Amodei published an open letter proposing that AI labs slow the pace of model-capability improvement. Three moves: embed third-party evaluators like METR inside labs, coordinate privately across competing labs under a US antitrust waiver, and align globally on narrow dangerous uses.
Sam Altman backed the plan publicly, saying pace-setting has been a primary topic at OpenAI for weeks. Elon Musk and Demis Hassabis also endorsed it on X within days.
But Nvidia CEO Jensen Huang pushed back at the 14 September All-In Summit, telling President Trump on speakerphone he wouldn't allow an AI slowdown. Trump called the movement a hoax.
Timeline
- 12 Sep 2026 — Dario Amodei publishes 'We must pace the frontier' letter; Altman, Musk, Hassabis endorse on X.
- 14 Sep 2026 — At All-In Summit, Nvidia CEO Jensen Huang tells Trump on speakerphone he won't allow a slowdown; Trump calls it a hoax.
- 15 Sep 2026 — OpenAI policy chief Chris Lehane confirms multi-week safety talks among OpenAI, Anthropic, Google; supports FRONTIER Act provision.
- 18 Sep 2026 — TechCrunch Equity podcast (34:22) breaks down what 'pacing the frontier' actually means in practice.
Why it matters for me
Altman, Musk and Hassabis backed the plan publicly and OpenAI policy chief Chris Lehane confirmed multi-week cross-lab talks; but Nvidia CEO Jensen Huang told President Trump he won't allow a slowdown.
Safety
Business Executives
If the three-lab coordination proceeds, US and global AI oversight could shift from post-hoc audits to embedded third-party verification inside frontier labs.
Daily Life · Privacy
General Public
Public trust in AI safety is under strain after researcher Jacob Coxon resigned over claims that labs are gambling with lives; the slowdown debate is a public answer to that loss.
Work
Ai Researchers
Frontier AI researchers face growing safety-pressure both internally (whistleblowers, evaluators) and externally (Trump/Huang backlash against slowdown), forcing a more visible ethics tradeoff.
What to remember
AI labs want to slow down. The US White House doesn't.
Amodei's slowdown plan has three AI-lab CEOs behind it and a Trump-Nvidia wall in front of it.
Verified sources (4)
Dario Amodei and other AI leaders want to 'Pace the Frontier' but...how?
↗Anthropic CEO outlines plan to slow AI development
↗Nvidia CEO Jensen Huang tells Trump 'we're not going to let [an AI slowdown] happen'
↗OpenAI, Anthropic, Google have been in talks on AI safety for weeks
↗Claims and linked sources
4 claimsAnthropic CEO Dario Amodei published a 12 September 2026 letter proposing three strategies to slow AI: embed third-party evaluators, cross-lab coordination, and global coordination on narrow dangerous uses.
Sam Altman, Elon Musk and Demis Hassabis publicly backed the slowdown proposal on X, and OpenAI policy chief Chris Lehane confirmed multi-week cross-lab safety talks including a FRONTIER Act provision.
At the 14 September All-In Summit, Nvidia CEO Jensen Huang told President Trump on speakerphone 'We're not going to let that happen' and Trump called the slowdown movement a hoax.
Reporter Brian Merchant criticised the plan, saying the concrete path from recursive AI improvement to mass harm remains undocumented, and warning the proposal risks serving Anthropic and OpenAI as regulatory capture.