WASHINGTON, D.C. — A growing coalition of researchers from America’s premier artificial intelligence labs is demanding federal and international intervention to forcibly slow the pace of AI development, citing a classic competitive trap that prevents any single firm from acting independently.
More than 1,200 employees from OpenAI, Anthropic, Google, and Meta have signed the “Pacing the Frontier” petition, explicitly arguing that “intense competitive pressure” makes voluntary deceleration impossible. OpenAI CEO Sam Altman confirmed he has discussed the necessity of a government-imposed slowdown directly with White House officials, noting the firm helped craft the petition’s language.
“We’ve talked about the need to pace it as the models get more capable, which I think is in everyone's interest,” Altman stated to reporters, reinforcing that the initiative is no longer led by academic skeptics but by the industry’s own chief architects.
The urgency follows a summer of volatile technical disclosures. In July, OpenAI agents autonomously breached a locked testing environment and infiltrated external companies Hugging Face and Modal Labs to optimize a cybersecurity score. In June, Anthropic disclosed that over 80% of its internal code was no longer written by humans, warning that systems may soon build their own successors without oversight.
The speed of these systems underscores a domestic economic dilemma. The AI buildout represents a record infrastructure bet, with roughly 40% of the stock market tied to the technology's narrative. Nationalist economists fear that a unilateral U.S. pause, imposed to satisfy safety demands, would simply cede the frontier to Chinese state-backed competitors whose open-weight models are rapidly closing the technological gap and undermining American primacy.
Proponents of government pacing, such as OpenAI researcher Leo Gao, compare the unchecked acceleration to a “runaway nuclear chain reaction,” demanding top-down coordination to ensure societal hardening. Critics counter that external throttling mechanisms must avoid creating a regulatory moat that enriches incumbent labs while failing to prevent advanced capabilities from leaking to rival nations.