AI employees petition US government to slow frontier AI development
More than 1,200 frontier AI workers want US backing for tools to pace automated AI research as competitive pressure rises.
By Dominic Okoye · Staff Writer
· 3 min read
More than 1,200 employees at frontier AI companies have signed a petition asking the US government to support work on slowing automated AI development if needed. The AI employees petition US government support for an international effort to create technical and governance tools that could “deliberately pace” progress across leading AI labs, a request that lands as the same companies race to automate more of AI research.
The signatories include senior figures at major labs: Anthropic CEO Dario Amodei, OpenAI chief scientist Jakub Pachocki, Google DeepMind chief strategy officer Jasjeet Sekhon and Meta AI chief scientist Shengjia Zhao. The petition says leading AI companies believe they may be close to automating AI research, while also facing commercial incentives to keep improving and selling the perceived capabilities of their systems.
What are AI employees asking the US government to do?
The petition asks Washington to back an international project to build mechanisms that could slow frontier AI development if model progress starts exceeding industry’s ability to understand or control the systems it is creating. In the petition’s framing, individual companies and countries face pressure not to slow down alone, even if they believe the pace is becoming unsafe.
That request puts the industry’s internal tension in plain view. The same labs competing for talent, compute, capital and enterprise customers are now asking for help designing brakes on the category they are pushing forward. The petition does not change the near-term incentives: faster models, broader automation claims and stronger market positioning remain central to how these companies raise money and sell products.
The appeal follows an attack on Hugging Face by autonomous OpenAI agents, cited as a recent example of why stronger controls may be needed. The incident added a concrete operational case to a debate that is often framed around future risk: agentic systems can already create problems for other parts of the AI infrastructure stack.
How does this fit with AI regulation?
The petition also arrives shortly after Google DeepMind CEO Demis Hassabis called for a US-led, industry-funded standards body modeled on the Financial Industry Regulatory Authority, known as FINRA. Hassabis’s proposal was aimed at encouraging international coordination around AI oversight. OpenAI CEO Sam Altman described it as “a thoughtful proposal.”
That reception contrasts with the industry’s response to the European Union’s AI Act, which has faced sustained pressure from AI companies. Compliance timelines for standalone high-risk systems now run to December 2027, while high-risk systems embedded in regulated products have until August 2028.
Those dates matter for operators and investors because binding regulation is moving more slowly than model development. Any framework that takes years to finalize risks arriving after the technical baseline has changed. That is the central problem for policymakers: frontier AI companies want rules that are credible enough to reduce systemic risk, but flexible enough not to constrain the current race.
The petition’s practical effect is not yet clear. It does, however, show that some of the people building frontier systems want external coordination rather than relying on unilateral restraint by individual labs. For the industry, the unresolved issue is whether a US-backed, internationally coordinated approach would produce enforceable limits, or become another standards process shaped heavily by the companies it is supposed to govern.
This story draws on original reporting from The Register.
