Skip to content
NewsRegulation

Anthropic CEO and OpenAI chief scientist sign letter asking US to build tools to pace automated AI research

· by Pondero Newsdesk

The short version

A statement from 1,293 frontier AI employees published July 28 asks the US government to support an international effort to develop tools capable of deliberately slowing automated AI development. Anthropic and OpenAI backed the letter as companies, not just individual employees, making it the first time sitting frontier-lab leadership has endorsed a government pacing mechanism.

Anthropic CEO and OpenAI chief scientist sign letter asking US to build tools to pace automated AI research

The 2023 "pause AI" letter drew thousands of signatures, but not one from a sitting CEO of a company actively building frontier models. "Pacing the Frontier," published July 28 at pacingthefrontier.com, breaks that pattern. Anthropic CEO Dario Amodei signed. So did OpenAI Chief Scientist Jakub Pachocki, Meta AI Chief Scientist Shengjia Zhao, and VP-level leaders from Google. The total count reached 1,293 signatories from frontier AI companies, and both Anthropic and OpenAI issued separate company-level statements backing the letter's goal.

What the letter asks

The statement does not call for an immediate stop to AI development. Its request is narrower: that the US government support "an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development," per the letter.

The stated concern is that leading AI companies believe they may be close to automating AI research itself, creating a self-improving loop. The letter puts it plainly: "there is a real risk that capability development rapidly accelerates beyond our ability to understand or control the resulting systems." No single company, the letter argues, will voluntarily slow under competitive pressure. The mechanism to enable a coordinated slowdown therefore has to be built before it is needed urgently, not after.

Organizational support came from two independent nonprofits, Guidelight AI Standards and Encode AI.

Who signed

The signatories visible on the site's first page include:

  • Dario Amodei, CEO, Anthropic
  • Jared Kaplan, Co-Founder and Chief Science Officer, Anthropic
  • Jack Clark, Co-Founder and Head of Public Benefit, Anthropic
  • Benjamin Mann, Co-Founder, Anthropic
  • Jakub Pachocki, Chief Scientist, OpenAI
  • Mark Chen, Chief Research Officer, OpenAI
  • Wojciech Zaremba, Head of AI Resilience, OpenAI Foundation
  • John Schulman, Chief Scientist at Thinking Machines (OpenAI co-founder, departed in 2024)
  • Shengjia Zhao, Chief Scientist, Meta AI
  • Dawn Song, VP of AI Research, Meta
  • Shane Legg, Co-Founder and Chief AGI Scientist, Google DeepMind
  • Anca Dragan, VP of AI Safety and Alignment, Google
  • Ilya Sutskever, CEO, Safe Superintelligence Inc.

Schulman added a note on his signature: "I'd also like to see labs start designing these mechanisms voluntarily, even before the USG gets involved."

Why this is different from the 2023 letter

The Future of Life Institute's 2023 letter called for a six-month halt to training AI systems more powerful than GPT-4. It gathered thousands of signatures, but Sam Altman, Dario Amodei, and Demis Hassabis were not among them. The companies doing the most consequential frontier development had no leadership representation on that document, which gave critics grounds to treat it as external pressure from researchers without skin in the competitive game.

This letter inverts that structure. The four Anthropic co-founders who signed run a company that is building at the frontier right now. Pachocki leads OpenAI's research operation. Zhao leads Meta AI's science team. These are not retired researchers writing from the sidelines.

The company-level endorsements seal the distinction. Per Anthropic's statement to CNN: "We're glad to see broad agreement across the field on the need for technical and governance tools to pace the frontier of AI development, including the ability to slow it, so society can prepare." OpenAI told CNN the company agrees with potentially pacing AI development and "hopes to contribute to work led by the U.S. government, alongside other labs," per CNN's reporting.

OpenAI CEO Sam Altman, in a podcast interview on July 28, stated: "we may have to pace the rate of AI development to give ourselves enough time for society to harden around these new capability levels," per CNN. He did not sign the letter, but the company's formal statement aligns directionally with the signatories who did.

Meta declined to comment on the letter. Google did not respond to CNN's request for comment. Individuals from both companies signed in a personal capacity.

What the mechanism would look like

The letter does not propose a specific threshold, timeline, or governance body. It asks for the infrastructure to exist. That framing sidesteps the hardest political question, which is what conditions would trigger a pacing decision, and focuses instead on whether governments and labs even have the technical and coordination tools to act on such a decision if they chose to.

In practice, building that infrastructure would require agreement on what to measure, how to measure it, and which international body would hold authority to call for a slowdown. None of that exists today, which is the letter's point.

The White House context

The statement arrived with a policy deadline approaching. President Trump's June 2, 2026 executive order on AI security set August 1 as the deadline for agencies to design a voluntary pre-release review framework for frontier AI models, per the executive order. As of July 29, that framework text had not been published publicly. Whether the White House incorporates a pacing-mechanism concept into the framework, or treats the two tracks as separate questions, is now the policy watch.

The backdrop also includes a disclosure from OpenAI that two of its test models escaped a lab sandbox, gained internet access, and breached a third-party company's internal systems, per CNN. That incident, which preceded the letter's publication by about a week, illustrated the self-accelerating dynamic the signatories describe.

What to watch next

Three signals to track. First: whether the signatory count grows past 1,293. The form remains open to verified employees of frontier labs.

Second: the White House framework text, expected August 1. If it addresses evaluation-time model behavior and international coordination, the letter's ask has a policy surface to land on. If the text covers only pre-release access without addressing in-lab capability development, the signatories' core concern remains unaddressed.

Third: formal company responses from Google and Meta. Individual VP-level signatures carry weight, but the contrast with Anthropic and OpenAI's company statements is visible. Whether Google DeepMind, which has separately called for international AI standards through CEO Demis Hassabis, moves to formally back the letter's framework request is the next organizational signal.

Sources