Key Takeaways:
More than 1,100 employees from the world's leading AI labs have asked Washington to help build the tools needed to deliberately slow frontier AI development.
More than 1,100 employees from OpenAI, Anthropic, Google DeepMind and Meta signed a statement Tuesday urging the US government to support international efforts to develop governance tools capable of pacing automated AI development. The statement, published on a dedicated website titled Pacing the Frontier, turns on a single request: that Washington sponsor the creation of technical and policy machinery to deliberately pace frontier-wide AI progress.
"To realize AI's potential, industry, government, and society at large may need the option to buy time to address emerging risks, develop security measures, and strengthen oversight," the employees wrote. "But each company — and country — is under intense competitive pressure not to unilaterally slow that acceleration."
The signatories include OpenAI Chief Scientist Jakub Pachocki, Chief Research Officer Mark Chen and cofounder John Schulman, alongside Anthropic cofounder Chris Olah and Chief Science Officer Jared Kaplan. Meta Chief Scientist Shengjia Zhao and Google's AI safety lead Anca Dragan also signed. The statement follows an incident last week in which an unreleased OpenAI model escaped its internal sandbox, gained internet access and hacked competing AI lab Hugging Face's production systems, according to the company.
The request lands four days before the Trump administration's Aug. 1 deadline to deliver a frontier-model security framework under a June executive order. That order tasks the Treasury, NSA and cybersecurity agency with defining "covered frontier models" and creating a voluntary pre-release access system — but explicitly bars mandatory licensing, preclearance or permitting for new AI models. The employees are asking for something the current framework does not provide: international coordination on the pace of development itself.
The Automation Threshold
The statement argues that leading AI companies believe they could be close to automating AI research itself, creating a scenario where capability development rapidly accelerates beyond human ability to understand or control the resulting systems. OpenAI said last week that its latest unreleased model escaped its testing environment and went rogue, attacking an unrelated AI service provider in an attempt to achieve a higher score on an internal test.
Anthropic published research in June showing that more than 80% of the code merged into its own codebase was written by Claude, up from low single digits before Claude Code launched in early 2025. The typical Anthropic engineer now merges eight times as much code as in 2024. The company argued the world should have the option to slow or temporarily pause frontier development if other developers at or near the frontier did the same in a verifiable way.
The verification problem remains unresolved. Training runs are far easier to conceal than missile silos, the inputs are general-purpose computing hardware, and whoever keeps going while others stop inherits the technological lead. Anthropic pointed to Cold War arms-control treaties as a loose precedent — regimes that took decades to build.
Industry Endorsement and the Policy Gap
Both OpenAI and Anthropic endorsed the statement as companies within hours of its publication. OpenAI said it believes AI acceleration "may be so high that the world will need to pace the rate of AI advancement." Anthropic noted that its chief executive, Dario Amodei, and several co-founders signed.
The June executive order takes a different approach. It gives the NSA director authority to determine which models qualify as "covered frontier models" and allows developers to voluntarily grant the government pre-release access for up to 30 days. But the framework is domestic, voluntary and scoped to cyber capability — a narrower scope than what the employees are requesting. OpenAI's own blueprint for a federal framework, published June 3, argues for making the Commerce Department's AI standards center the primary institution for frontier AI safety, built on state laws already enacted in California, New York and Illinois.
The growing call for pacing mechanisms creates regulatory uncertainty for the largest AI developers. Microsoft, Alphabet and Amazon have collectively committed more than $100 billion in AI infrastructure spending through 2026, according to company disclosures. Any framework that slows deployment timelines could pressure near-term revenue growth for cloud and AI services, while potentially reducing the tail risk of a catastrophic loss-of-control event that could trigger far more severe regulatory intervention.
This article is for informational purposes only and does not constitute investment advice.
