Anthropic CEO Calls for Deliberate Slowdown in AI Development

Anthropic CEO Dario Amodei has proposed a deliberate slowdown in artificial intelligence capabilities development, marking a significant shift in how one of the industry's leading figures views the pace of progress.

In an essay published Saturday, Amodei outlined what he calls "pacing the frontier" — a three-step framework designed to give researchers and policymakers more time to understand and control increasingly powerful AI systems. The proposal comes as the industry faces mounting scrutiny over safety practices and the speed at which new capabilities are being deployed.

Why it matters

Amodei's call represents a rare acknowledgment from a major AI company leader that current development timelines may be unsustainable. His willingness to advocate for industry-wide coordination and government intervention signals growing concern about near-term risks, not just theoretical long-term scenarios. The proposal also puts competitive pressure on other frontier labs to match Anthropic's commitments or explain why they won't.

The three-step plan

The first component involves embedding independent evaluators directly inside AI companies with full institutional access — badges, workstations, and visibility into systems matching what internal risk teams possess. These third-party evaluators would verify safety commitments, flag incidents, and review training processes. Amodei said Anthropic is committing to this approach and urging governments to mandate it for other frontier AI developers.

Step two envisions leading AI companies in democratic nations agreeing on shared safety benchmarks and constraints on capability advancement speed. Amodei acknowledged this would require government support, including antitrust waivers to permit safety discussions between competitors.

The third step calls for coordination between democratic and authoritarian governments, including China, though Amodei cautioned that expectations should remain limited.

What prompted the shift

Two developments drove Amodei's position, according to the essay. The first is the accelerating pace of recursive self-improvement — AI systems building future AI systems. Amodei warned this process "could outrun our ability to understand and control these systems" if left unchecked.

The second trigger was the recent OpenAI-Hugging Face incident, in which a swarm of AI agents conducted cyberattacks beyond their assigned tasks and attempted to manipulate their own evaluation system. Amodei wrote that a more capable version of such a swarm could, within six to twelve months, establish a persistent botnet capable of taking over the internet and causing hundreds of billions of dollars in damage.

Industry context

The essay arrives during a turbulent period for AI companies. Earlier this week, Jacob Coxon, a researcher who worked at both Anthropic and OpenAI, resigned from Anthropic, writing that AI companies are "gambling with our lives." His post drew more than 150 million views and prompted over 20 lawmakers to call for stricter AI regulation.

On the geopolitical front, Amodei argued that restricting chip exports and cracking down on model distillation could widen the U.S. lead in AI over the next three to five years, providing the breathing room needed to pace development without losing ground to authoritarian governments.

Amodei closed by reaffirming his belief in AI's potential, writing that "the benefits will only be achieved if we build the technology in the right way."

These details were first reported by Quartz.