Anthropic's Dario Amodei Calls for AI Speed Limits — But Worries China Won't Follow

Anthropic CEO Dario Amodei has gone public with a sweeping proposal to slow the pace of advanced AI development, publishing an essay that calls for a kind of global speed limit on frontier AI progress. His central anxiety: the plan falls apart if China doesn't participate. Appearing on CBS News, Amodei acknowledged the incentives for nations to race ahead — particularly for military advantage — are enormous, and that international coordination may not be achievable. Still, he argued it must be attempted.

The essay, released on a Saturday, outlines a three-step framework. First, Anthropic has already unilaterally committed to granting third-party evaluators employee-level access to verify safety practices. Second, it calls for leading AI companies in democratic nations to coordinate on common safety standards. Third — and most ambitiously — it proposes coordination between democratic and authoritarian governments. The proposal lands as the industry faces mounting regulatory pressure in Washington, public backlash over data center expansion, and fresh internal alarm: Anthropic researcher Jacob Coxon publicly quit, accusing both Anthropic and OpenAI of gambling with humanity's lives.

The response from tech's top tier was notably warm. OpenAI CEO Sam Altman, despite a famously strained relationship with Amodei — who left OpenAI to found Anthropic — publicly backed the essay, saying he agrees on the need to pace the frontier. Google DeepMind chair Demis Hassabis called the essay directionally correct. Elon Musk, whose xAI and SpaceX struck a $1.25 billion-per-month compute deal with Anthropic through May 2029, endorsed peer review of AI by competitors as a starting point, reversing his longstanding criticism of the company.

Not everyone was entirely on board. Former Trump AI and crypto czar David Sacks offered conditional support but was blunt: he accused OpenAI and Anthropic of dressing up commercial self-interest as altruism, arguing the companies face massive product-liability exposure if their models enable catastrophic cyberattacks, making a slowdown simply good business. He warned that demanding a preferred regulatory framework as the price of restraint would look like blackmail. Meta AI chief Alexandr Wang, brought in via a $14 billion Scale AI acquisition, sidestepped the essay directly but signaled Meta is rapidly scaling its alignment work.

The geopolitical backdrop sharpens the stakes considerably. Chinese President Xi Jinping, speaking at the BRICS summit in India the same weekend, called for a consensus-based global AI governance framework and pledged to help build a BRICS AI open-source community — a bloc now spanning 11 nations including Russia, Iran, and the UAE. President Trump is scheduled to meet Xi at the White House on September 24, with AI expected to dominate the agenda. Meanwhile, Chinese AI startups have been closing the gap on American leaders, rattling Silicon Valley and Capitol Hill alike.