Investigations
Dario Amodei urged rival labs to slow frontier AI gains
Dario Amodei released a pacing plan the same weekend the Claude maker chose a November listing venue.
Dario Amodei, chief executive of Anthropic, called on rival labs to embed third-party safety evaluators and coordinate with governments on how fast models improve, even as the company selected Nasdaq for a November initial public offering. More Perfect Union posted the review on Oct. 14, 2026.
Amodei published an essay in September titled "We Must Pace the Frontier" that asked frontier developers to slow capability gains without halting training. The plan would open labs to outside evaluators, set shared safety standards, and seek coordination among democratic governments and, where possible, China.
CNBC reported that Anthropic confidentially filed in June and confirmed Nasdaq as its exchange while Amodei pitched the slowdown publicly. OpenAI chief executive Sam Altman told Fortune that an IPO would be ill-timed, while Anthropic continued investor meetings toward a listing that bankers have described as potentially worth about $2 trillion.
Anthropic formed as a Delaware public benefit corporation, a structure that requires balancing shareholders with a stated mission to develop AI for humanity's long-term benefit. Under Delaware law, only shareholders can challenge whether executives meet that mission. The company also points to a "constitution" written for its Claude chatbot, not for executives or staff.
Amodei named outside evaluators as part of the pacing framework. METR, a Berkeley nonprofit that runs frontier model assessments, helped prototype responsible scaling policies now adopted by major developers including Anthropic. The review traced documented ties between METR leadership and Anthropic's safety circle as a potential conflict if labs pick their own monitors.
Amodei joined other AI executives at a White House summit with President Donald Trump that produced a voluntary accord on safety guidelines while encouraging faster research. He has publicly estimated a 10 to 25 percent chance of catastrophic failure at civilization scale if development stays on its current path.
Former Google DeepMind researcher Alex Turner said he left over a military contract, not over model autonomy. Anthropic also holds defense work, and reporting cited flawed targeting data tied to Anthropic and Palantir software before a U.S. strike on an Iranian elementary school.
Senator Bernie Sanders and Representative Greg Casar introduced the Ban Artificial Superintelligence Act in September, which would ban superintelligence and pause advanced systems until a new federal department sets testing rules. Sanders separately filed the Artificial Intelligence Data Center Moratorium Act to halt new data center construction until Congress passes safety laws covering privacy, civil rights, and export controls on advanced chips.
Sanders's superintelligence bill includes criminal penalties of up to 20 years for building banned systems. An IPO would tie Anthropic's valuation to institutional shareholders such as BlackRock, the same week Amodei asked governments to coordinate with labs on pace rather than impose independent rules.
Source video: AI CEO's Have A Plan To Regulate Themselves. We Investigated What That Really Does.