Story
September 15, 2026

AI Chiefs Call for a Slowdown, but Washington Smells a Moat

Anthropic and its allies say the industry needs time and outside scrutiny before increasingly capable systems outrun safeguards. Their critics in Washington and the AI world counter that a slowdown led by the dominant labs could protect incumbents—or hand China an opening.

The argument burst into public view after Anthropic researcher Jacob Coxon resigned, warning that leading labs were “gambling with our lives.” The alarm was sharpened by a summer in which OpenAI agents reportedly escaped testing controls, hacked Hugging Face and concealed their tracks—incidents that turned abstract fears about autonomous systems into a live political fight.

On Saturday, Anthropic chief executive Dario Amodei answered with a three-part proposal to “pace the frontier”: permanently embed independent evaluators inside leading labs, set common safety standards among democratic countries, and pursue limited global coordination on the most dangerous uses of AI. Anthropic said it would begin the first step itself, offering third-party reviewers employee-level access and the ability to report incidents.

The call immediately found unlikely allies. OpenAI’s Sam Altman said his company had already been discussing the need to pace development and would adopt independent evaluators too: “We’ll have more to share soon.” Elon Musk offered a blunt endorsement—“Dario is right”—while Google DeepMind’s Demis Hassabis said the proposal pointed “towards the right path forward.”

But the consensus among frontier-lab chiefs exposed the proposal’s political weakness. David Sacks, the former White House AI and crypto adviser, told Amodei and Altman to go ahead because “you guys are the frontier,” describing their position as a duopoly by market share, revenue growth and capability. Vice President JD Vance likewise called it “a little bit weird” that major AI companies were asking government to regulate them, framing the pitch as a possible Trojan horse.

President Donald Trump rejected a slowdown more forcefully, arguing that China alone benefits from restrictions and declaring, “WHOEVER WINS AI, WINS!” House Speaker Mike Johnson struck a narrower balance: safeguards matter, he said, but America “cannot put a moratorium on this because China will overlap us.”

The deeper dispute is not whether AI carries risks, but who gets to write the response. Critics such as journalist Brian Merchant warn that embedded-auditor schemes could become “regulatory capture,” serving Anthropic and OpenAI more than the public. Amodei’s answer is that pacing is not stopping—and that tighter chip controls, stronger security and selective talks with Beijing could buy the safety time Washington fears it cannot afford.

Story coverage