Anthropic CEO Proposes Three-Step Plan to Slow AI Development

Anthropic CEO Dario Amodei has published an essay calling on artificial intelligence companies to deliberately slow the pace at which they advance model capabilities, proposing a three-step framework he says can temper development without sacrificing competitive position.

The essay, published Saturday and first reported by CNBC, arrives amid heightened scrutiny of AI safety practices and follows a high-profile resignation from Anthropic's own research team earlier this week.

The three-step framework

Amodei's plan begins with third-party verification. Anthropic has already committed to this first step, granting outside evaluators employee-level access to verify safety practices and report incidents. The company characterized this as a "unilateral" commitment.

The second step calls for leading AI companies in democratic countries to coordinate and establish common safety standards. The third envisions coordination between democratic and authoritarian governments, though Amodei acknowledged some steps will prove harder to implement than others.

"To be clear, pacing does not mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models, and for third party evaluators to confirm this," Amodei wrote.

Why it matters

The proposal represents a significant shift in tone from a CEO whose company is preparing for what observers expect to be a major IPO. By publicly advocating for industry-wide slowdowns, Amodei is positioning Anthropic as willing to accept constraints on development speed—a stance that could influence regulatory discussions and competitive dynamics as AI capabilities approach levels that researchers consider potentially dangerous. The timing also suggests internal and external pressure on AI labs is mounting.

Researcher departure sparks debate

The essay followed the resignation of Jacob Coxon, an Anthropic researcher who previously worked at OpenAI. Coxon announced his departure on social media, saying he left because Anthropic and OpenAI are "gambling with our lives." He warned that people building AI systems "earnestly believe that it could kill us all by the end of the decade."

While such warnings may sound extreme, concerns about existential risk from AI have circulated in research communities for years. In 2023, prominent figures including OpenAI CEO Sam Altman and Amodei himself signed a statement declaring that "mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war."

Timing and rationale

Amodei explained that calls to pause or slow AI development have existed since 2023, but made "little sense" at that time because models lacked the power to take real-world action and were not yet capable of "significant deception, manipulation, cheating, or cyberattacks."

The implication is that current or near-future systems may cross those thresholds, making coordinated pacing more urgent.

"I continue to believe that AI can enormously improve the quality of human life," Amodei wrote. "But the benefits will only be achieved if we build the technology in the right way, and—so long as we use the time we gain well—it is worth taking unusually deliberate care to get it right."

Details were first reported by CNBC.