AI leaders sign statement asking the government to do something about automated AI
Employees of OpenAI and Anthropic, as well as Google, Meta, Thinking Machines, Microsoft, Mistral, and other leading AI labs, have written a statement to the US government supporting a potential slowdown of sorts for frontier AI development - or at least a speed-up of global coordinated governance efforts.
A letter signed by employees of OpenAI, Anthropic, Google, Meta, Thinking Machines, Microsoft, and Mistral has asked the United States government to consider slowing the frontier of artificial intelligence development, or at minimum to accelerate coordinated global governance. The signatories frame the request in cautious language, warning that the world's leading AI companies may be close to automating AI research itself, and that the resulting acceleration could outpace any meaningful oversight.
The mechanics here are worth noting. The signatories are not executives. They are engineers and researchers, the people closest to the technical frontier, and they have chosen to address Washington rather than their own boards. That choice reveals where the leverage sits. Inside these labs, the people who understand the systems have concluded that internal governance is insufficient, and that the only counterweight capable of slowing the pace is external.
The phrase "automating AI research" is the load-bearing claim. If a model can improve the next model, the development curve stops being linear and becomes recursive. Recursive improvement compresses the time available for regulators, competitors, and society to react. The signatories are not arguing against AI. They are arguing that the gap between capability and control is closing faster than the institutions built to manage it.
For the labor market, the implications are direct. A slowdown in frontier development, if it occurs, would preserve a window in which human researchers, engineers, and adjacent knowledge workers remain the primary drivers of progress. A coordinated governance regime would likely formalize disclosure requirements, safety evaluations, and deployment thresholds, all of which create new categories of compliance and oversight work. The remote-first firms that already operate across jurisdictions would inherit the most complex version of that compliance burden, and the most opportunity from solving it.
The statement is also a signal about the internal politics of these labs. Employees publicly disagreeing with the pace set by their employers is a rare event in an industry that prizes secrecy. It suggests that the concern is not fringe. It is shared by people who have seen the systems, understand the trajectory, and decided that the cost of speaking up is lower than the cost of staying silent.