More than 1,100 employees across the frontier AI industry put their names this week to a statement calling on the US government to help build the technical and policy tools needed to deliberately slow the pace of AI progress. Signatories named by Axios on July 28, 2026, include OpenAI chief scientist Jakub Pachocki and Anthropic co-founder Jared Kaplan, two of the most senior technical figures inside their respective labs.
The statement, hosted at pacingthefrontier.com, does not ask any lab to change a release schedule or hold back a model. It asks Washington to support an international push that would give the industry room to close safety gaps, harden security practices, and build up regulatory oversight, an option the signatories say does not currently exist because no single company can slow down without ceding ground to rivals.
Underneath that ask sits a striking admission: engineers inside the very labs racing to build frontier systems think their own employers may soon be capable of automating AI research itself. The signatories argue that once that happens, capability could race ahead of the field’s own ability to understand or safely manage what it has built, and that the tools, technical and regulatory, needed to consciously pace that progress simply do not exist yet.
What the letter does not do is commit OpenAI, Anthropic, or any other employer to a policy position. Pachocki and Kaplan signed as individuals. Neither company has said it will change how it operates, and the statement itself is addressed outward, at regulators and lawmakers, rather than inward, at the executives who set each lab’s roadmap.
That is a defensible division of labor rather than a dodge. A single company choosing to slow down while competitors do not simply hands the frontier to whoever moves fastest, which is exactly the collective action problem the statement describes. Building the shared monitoring and governance capacity it calls for is, by design, a job for governments and international bodies, not for any one lab acting alone.
The statement is also short on specifics. It nods to existing efforts that already track which frontier models get released, without naming a reporting threshold, a compute limit, or a licensing scheme for Washington to adopt.
The open question for the next few months is whether Congress or the administration treats this as a mandate to build real pacing infrastructure, or lets a list of 1,100 names function as a talking point with no legislative follow-through.
This account draws on the statement published at pacingthefrontier.com and on Axios’s July 28, 2026 report identifying signatories including Jakub Pachocki and Jared Kaplan.
