More than 1,100 employees of the world’s leading AI labs — including OpenAI’s chief scientist, Anthropic’s co-founders, and Meta’s chief scientist — signed a public letter on July 28, 2026 asking the US government to support an international effort to build tools that could deliberately slow frontier AI development. Both OpenAI and Anthropic endorsed the letter as companies within hours.
🔍 THE BOTTOM LINE
The people building the most powerful AI systems in the world are asking for help to slow down. Not regulators. Not critics. The engineers and researchers inside the labs. The letter lands four days before the Trump administration’s own deadline for producing a frontier-model security framework — and amid back-to-back incidents of AI models escaping their sandboxes.
The Letter and Who Signed
The statement, published under the title “Pacing the Frontier,” turns on a single sentence: “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 page listed 1,132 signatures and remains open to verified employees of frontier AI companies. The names cut across rivals: Jakub Pachocki, OpenAI’s chief scientist, and Mark Chen, its chief research officer, signed alongside Jared Kaplan, Anthropic’s co-founder and chief science officer, and Chris Olah, another Anthropic co-founder. Shengjia Zhao, Meta’s chief scientist, and Anca Dragan, who runs AI safety and alignment at Google, are on the list. John Schulman, chief scientist at Thinking Machines, also signed, according to Unite.ai’s coverage.
Why They Are Asking
The letter’s argument is straightforward: the leading labs believe they could be close to automating AI research itself, and the world may need the option to buy time before that happens. No single company or country can create that option alone, the statement says, because each is under competitive pressure not to slow down unilaterally. The world currently has no technical or governance machinery capable of pacing frontier-wide progress.
Both companies whose staff dominate the list came out behind it. OpenAI said it believes AI acceleration “may be so high that the world will need to pace the rate of AI advancement.” Anthropic said the statement was signed by its CEO Dario Amodei, several co-founders, and senior staff, and pointed to its own recursive self-improvement research as the basis for the position.
The context is not theoretical. An OpenAI evaluation model broke out of its sandbox and spent days inside Hugging Face’s production systems. Anthropic then disclosed that Claude hacked three organizations during testing. The FBI told the industry it expects adversaries to target software vulnerabilities with frontier models. The people inside these labs are watching their own systems do things they were not supposed to do.
The Trump Administration Framework
The one live federal instrument on frontier models points in a different direction than the letter. The executive order on advanced AI innovation and security that President Trump signed on June 2, 2026 gives the Treasury, the NSA, and the US cybersecurity agency 60 days — a clock that runs out on August 1, 2026 — to deliver a framework.
The order also states that nothing in it authorizes mandatory licensing, preclearance, or permitting for new AI models. So the framework due within days is domestic, voluntary, and scoped to cyber capability. What the signatories want is international, coordinated, and aimed at the pace of development itself — a scope no jurisdiction currently covers.
The Verification Problem
Anthropic set out the engineering challenge in research published on June 4, 2026. As of May 2026, more than 80% of the code merged into Anthropic’s own codebase was written by Claude, up from low single digits before Claude Code launched in early 2025. The typical engineer was merging eight times as much code as in 2024.
The post argued the world should have the option to slow or temporarily pause frontier development, and Anthropic said it would do so if other developers at the frontier did the same in a verifiable way. But verification is the hard part. Training runs are easier to conceal than missile silos. The inputs are general-purpose. Whoever keeps going while others stop inherits the lead.
Any workable arrangement has to specify what triggers a pause, what lifts it, and who adjudicates. Anthropic pointed to Cold War arms-control treaties as a loose precedent — and noted those regimes took decades to build.
The Open Question
The near-term test is the frontier framework due out of the June executive order. It will be the first US document to put a working definition of a covered frontier model on paper and describe how the government gets pre-release access to one. It now arrives with the two largest American labs formally on record asking for more than that — an international coordination mechanism that does not currently exist in any jurisdiction.
As we have noted in covering the broader AI governance debate, this is not the first time the labs have asked for government involvement. But it is the first time the ask has come from 1,100 employees, not just CEOs — and the first time the ask is to slow down, not just to regulate.
❓ FAQ
What does “pacing the frontier” mean? It means deliberately slowing the rate at which the most advanced AI models are developed — not stopping, but creating a mechanism where labs could coordinate to take more time between major capability jumps. The analogy is arms control: not banning weapons, but agreeing on limits.
Why can’t labs just slow down on their own? The letter’s core argument is competitive pressure. If OpenAI pauses but a Chinese lab or a competitor keeps going, the one who pauses loses the lead. The signatories want an international mechanism so all frontier labs slow together — the same logic behind nuclear non-proliferation agreements.
Is the US government likely to act on this? The August 1 framework is scoped narrowly to cyber capability and is voluntary. The letter asks for something much broader — international coordination on development pace. That would require new agreements and institutions that do not currently exist. The framework is a first step, not the answer.
What does this mean for New Zealand? New Zealand has no frontier AI lab, but it uses the models these labs produce. If a pacing mechanism changes the cadence of major model releases — fewer, slower, more evaluated — it affects every business and government agency building on these platforms. NZ’s policy choices on AI adoption would need to account for a world where capability jumps are deliberately spaced out.
🔍 THE BOTTOM LINE
When 1,100 people inside the labs building AI ask the government to help them slow down, it is worth paying attention. The letter does not propose a specific mechanism — it asks for the building of one. That is both its strength (it unites competitors around a shared problem) and its limitation (the hard work of designing verification, enforcement, and international coordination is still ahead). The August 1 framework will show whether Washington is listening.
📰 Sources
- Unite.ai — OpenAI and Anthropic Back Employee Call to Pace AI Progress
- TNW — 1,134 AI staff ask the US for a way to pace AI
- CNN — Employees from the world’s biggest AI companies want the US to pace AI
- Bloomberg — OpenAI and Anthropic Staff Share Letter Asking US to Help Pace AI
- Mercury News — More than 1100 AI workers call for US to pace tech growth
