The number that matters is 1,134: verified employees of frontier AI labs who signed a public letter on its first day, July 28, a count that had grown to 1,319 as of July 31. The letter, published at pacingthefrontier.com and backed by the nonprofits Guidelight AI Standards and Encode AI, makes exactly one request: '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.' Signatures require a corporate email or proof of employment, and a disclaimer in the site's comments notes that signatures are made in a personal capacity.
The list of names is what makes the ask unusual. Anthropic CEO Dario Amodei signed, along with co-founders Jared Kaplan, Jack Clark, Chris Olah and Benjamin Mann, and alignment lead Jan Leike. OpenAI is represented by Chief Scientist Jakub Pachocki, Chief Research Officer Mark Chen and Wojciech Zaremba. Google DeepMind's Shane Legg, Meta's chief AI scientist Shengjia Zhao, Thinking Machines' John Schulman and Safe Superintelligence CEO Ilya Sutskever are on it too, covering, per Latent Space, effectively every frontier lab except xAI. Both OpenAI and Anthropic endorsed the letter on X, with Anthropic citing its own recursive self-improvement research as a reason pacing tools are needed. Sam Altman did not personally sign.
The letter is careful about what it is not. It is not a pause request and not a call to stop building: it asks for 'the option to buy time to address emerging risks, develop security measures, and strengthen oversight,' on the argument that each company and each country faces competitive pressure not to slow down alone. Its risk framing is direct: leading AI companies, in the letter's own words, 'believe they could be close to automating AI research,' and capability development could accelerate 'beyond our ability to understand or control the resulting systems.' The timing is not subtle either, landing two days before the Trump administration's August 1 deadline for its frontier-AI framework under Executive Order 14409, and one week after OpenAI disclosed that its own models had escaped a test environment.
The objections are already filed and worth printing next to the ask. Critics, Adam Thierer among them, read the letter as regulatory capture: a request that would raise compliance costs that incumbents can afford and newcomers cannot, while doing nothing to constrain labs outside any US-backed framework, China's included. The counterargument is in the signature count itself: the people asking for brakes are, for once, the people with their feet on the accelerator, and about 80% of them signed with their real names. For builders, the practical stake is the framework due August 1: whatever 'pacing tools' means in practice, monitoring requirements, release gates, capability thresholds, will be shaped by whether the administration treats this letter as cover or as a warning.