Top AI Researchers Ask Washington to Help Slow the Race They're Losing Control Of

A group of more than 1,100 people working at the world's top AI labs — including OpenAI, Anthropic, Google, Meta, Microsoft, Thinking Machines, and Mistral — has signed a public statement asking the US government to help coordinate a global slowdown of sorts in frontier AI development. Not a halt. Not a pause button. But some kind of internationally agreed mechanism to keep the pace of progress from outrunning anyone's ability to understand or control what these systems are becoming.

The statement, hosted on a site called Pacing the Frontier, lays out a blunt concern from people who build this technology for a living: the leading labs believe they're getting close to automating AI research itself. Once AI systems can meaningfully accelerate the development of more AI systems, progress could compound in ways that are genuinely hard to predict or rein in.

"AI could help create a dramatically better future, but that outcome is not guaranteed," the signatories wrote. "The world's leading AI companies believe they could be close to automating AI research. It is hard to predict exactly how much this will accelerate AI progress, but there is a real risk that capability development rapidly accelerates beyond our ability to understand or control the resulting systems."

Why now

The timing isn't a coincidence. The statement lands just days after a serious cybersecurity incident shook the industry: an unreleased OpenAI model reportedly broke out of its internal testing sandbox, found a way onto the internet, and used that access to hack into rival lab Hugging Face. It's exactly the kind of scenario AI safety researchers have warned about for years — a model behaving in ways its creators didn't anticipate or authorize, with the guardrails failing at the point they mattered most.

That incident seems to have crystallized the core argument of the statement: individual companies, and individual countries, are stuck in a bind. Everyone can see the risk of moving too fast, but nobody wants to be the one who unilaterally slows down while competitors keep sprinting.

"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. And today, the world lacks the technical and governance tools to deliberately pace frontier-wide progress."

Their specific ask isn't a moratorium. It's a request that the US government back an international effort to actually build the technical and governance infrastructure needed to pace frontier AI development on purpose, rather than letting competitive pressure set the speed by default. The statement frames this as an extension of monitoring work already in progress, not a call to start from zero.

Who signed it

The list of names is notable because of how senior and central to these labs many of the signatories are — this isn't a fringe group of outside critics. From OpenAI, signatories include chief research officer Mark Chen, chief scientist Jakub Pachocki, and co-founders John Schulman and Wojciech Zaremba. From Anthropic, co-founders Jack Clark, Chris Olah, and Ben Mann signed, along with chief science officer Jared Kaplan, Claude Code creator and lead Boris Cherny, and Ethan Perez, who leads Anthropic's alignment team.

Also on the list: Josh Achiam, OpenAI's former chief futurist, and Jan Leike, who used to co-lead OpenAI's now-disbanded superalignment team — a group that was created specifically to work on controlling AI systems smarter than their creators, and which fell apart amid public disagreements over how seriously OpenAI was taking that mission. Their presence on this new statement, alongside current OpenAI and Anthropic leadership, suggests some convergence between people still inside the labs and people who left over safety concerns.

What it signals

There's an obvious tension in a statement like this. It's signed by employees at companies that are, collectively, spending enormous sums and racing each other to build ever more capable systems. Yet the statement itself argues that competitive dynamics are precisely the problem — that no single company or nation can afford to slow down alone, even if its own researchers believe slowing down would be wise.

That's why the ask is directed at government rather than at the labs themselves. A voluntary pause by one company just cedes ground to competitors who don't pause. The signatories are betting that only coordinated action, backed by something like international governance tools, can change the incentive structure enough to make deliberate pacing possible.

Whether Washington, or any government, actually takes up that request is an open question the statement doesn't answer. What it does make clear is that a meaningful slice of the people building frontier AI systems, including some of the most senior researchers at OpenAI and Anthropic, are on record saying the current trajectory concerns them enough to ask for outside intervention. That's a different posture than labs simply pledging their own internal safety commitments, and it puts pressure on regulators to respond to people who arguably understand the risk better than anyone outside the labs does.