Wednesday, Jul 29 | --:--
Back to home

1,200+ Frontier AI Employees Ask U.S. to Back Tools That Can Pace Automated AI R&D

On July 28, 2026, more than 1,200 employees at frontier AI companies published “Pacing the Frontier,” urging the U.S. government to support international technical and governance tools to deliberately pace automated AI research—backed the same day by OpenAI and Anthropic as organizations.

Tech Insights Reporter 5 min read Washington, DC
Cover illustration for 1,200+ Frontier AI Employees Ask U.S. to Back Tools That Can Pace Automated AI R&D

TLDR

On July 28, 2026, “Pacing the Frontier”—a public statement from 1,224 employees of frontier AI companies—asked the U.S. government to support an international effort to build technical and governance tools that could deliberately pace the frontier of automated AI development. Signatories include Anthropic CEO Dario Amodei, OpenAI Chief Scientist Jakub Pachocki, Meta Chief Scientist Shengjia Zhao, and leaders across Google / DeepMind. OpenAI and Anthropic posted company-level support the same day. The petition is not a concrete kill-switch bill; it is a request for option value—the ability to buy time if AI research automation races ahead of control.

What the statement says

Primary site: pacingthefrontier.com. Core claim structure:

  1. Leading labs may be close to automating AI research.
  2. Acceleration could outrun society’s ability to understand or control systems.
  3. Competitive pressure blocks unilateral slowdowns by any company or country.
  4. Therefore: “We request that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development.”

The text is short by design. It does not specify thresholds, enforcement design, China strategy, open-source treatment, or compute controls. Signers emphasize establishing common knowledge that coordination tools may be needed—not a freeze tomorrow.

Who signed and who endorsed

Layer Details (primary + wire)
Employee statement 1,224 frontier-company employees (count on primary site)
Named senior signatories (examples) Dario Amodei, Jared Kaplan, Jack Clark, Benjamin Mann (Anthropic); Jakub Pachocki, Mark Chen, Wojciech Zaremba (OpenAI); Shengjia Zhao, Dawn Song (Meta); Anca Dragan, Laura Weidinger, Jasjeet Sekhon (Google / DeepMind); John Schulman (Thinking Machines)
Company support Anthropic (@AnthropicAI): supports the petition; CEO, co-founders, senior staff signed; points to recursive self-improvement research
Company support OpenAI (@OpenAI): mission framing that frontier acceleration may eventually require pacing tools led by the U.S. government with labs and open-source community

Nonprofits Guidelight AI Standards and Encode AI are listed in Reuters-style coverage as supporting the initiative’s organization.

Context

The petition lands days after OpenAI’s Hugging Face cyber-eval incident disclosure and the House AI kill-switch / DHS authority bill story—same policy weather, different instrument. It is staff-driven, not a board resolution; Meta and Google leaders appear as individual signers while company product strategies continue at full speed.

Why this story matters

When chief scientists and CEOs of rival labs co-sign a request for pacing tools, the Overton window for international coordination moves even if the policy design is blank. The next fight is operational: who holds the tools, what triggers them, and whether “pace” means evaluate-before-release, compute caps, or something closer to an emergency throttle. Watch Congress (FRONTIER Act–style bills), agency R&D on monitoring, and whether labs start voluntary pacing mechanisms before any statute.

Sources

Prior Coverage

Earlier Times of AI reporting on this thread.

Scroll to continue reading