One of the loudest boosters of cutting-edge chatbots is now arguing the labs behind them should tap the brakes—and his rivals instantly agreed. In a lengthy essay posted on his website, Anthropic CEO Dario Amodei urged a global deceleration in AI progress, saying that safety work is falling behind the technology's capabilities and that labs should "pace the rate of capabilities advancement" so guardrails can catch up, the New York Times reports. In one example of what could happen otherwise, per the AP, Amodei warned that AI could take over the entire internet in less than a year.
His call lands days after an Anthropic researcher, Jacob Coxon, publicly quit over safety concerns. Amodei's essay proposes third-party "embedded evaluators" with deep access inside labs and coordinated safety standards among democratic countries, potentially through regulation, while acknowledging any real slowdown would have to be global, including authoritarian states. Analysts aren't convinced that these statements are purely about public safety, arguing that big labs are using such warnings to shape rules that entrench their dominance and sell new security products, per the Times.
Amodei's competitors quickly fell in line on Saturday. "Dario is right," Elon Musk posted on X. The independent evaluators proposal is "a great idea," OpenAI CEO Sam Altman posted, per Reuters, saying, "I agree with Dario that we need to pace the frontier." Amodei wrote that the upside of AI remains "incredible," mentioning cures for diseases and faster economic growth, it's just that the industry owes it to humanity to move forward more cautiously. The downside, three Anthropic staff members have said, would be the extinguishing of the human race.
,
uBlock 
,
and
Ghostery 
)
