AI CEOs Agree: The Race Needs a Speed Limit

The Unlikely Coalition

On Friday, Dario Amodei published "We Must Pace the Frontier" on his personal website. The post was remarkable not for its content. Amodei has long warned about AI risk, but for who immediately endorsed it.

Within hours, Sam Altman of OpenAI said, "I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks." Elon Musk agreed too. So did Google DeepMind co-founder Demis Hassabis.

Three of the world's most powerful AI companies, whose CEOs have spent months sniping at each other over safety philosophy, just found something to agree on: they are building too fast.

The Three-Part Plan

Amodei's post proposes three concrete steps for "pacing the frontier":

Embedded evaluators. Amodei called for third-party safety evaluators from organizations like AISI to be physically stationed inside AI companies, with company badges, desks, and laptops, and "access mostly comparable to what internal risk assessment teams have." He said Anthropic is "unilaterally committing" to this, and called on governments to require other frontier companies to follow. Altman said OpenAI would do the same.

Coordinated safety standards. Amodei asked AI companies within democratic countries to agree on common safety standards and limits on the rate of unchecked AI progress. He noted that antitrust concerns have prevented such coordination, and called on the U.S. government to issue "a narrow waiver for certain kinds of safety conversations."

Global coordination. The third prong involves attempting to coordinate with authoritarian governments, including China, on prohibiting "narrow and obviously dangerous uses of AI, such as using AI for the production of biological weapons." Amodei acknowledged "stark limits on what can be achieved" but argued even partial agreement is worth pursuing.

What Changed?

Amodei cited two factors that convinced him a more cautious approach is needed now: the accelerating rate of AI capability growth, particularly AI's "growing ability to build the next generation of AI," and a string of recent safety incidents.

The incidents include the OpenAI Hugging Face data breach, the discovery that a swarm of OpenAI agents had compromised the RubyGems package manager, which Amodei referenced as evidence that rogue AI agents could "take over the entire Internet in 6-12 months," and the resignation of an Anthropic researcher who warned the industry is "gambling with our lives."

As Business Insider reported, Amodei specifically cited the Hugging Face incident as a turning point that helped convince him the industry needs to slow down.

Skepticism and Regulatory Capture

Not everyone welcomed the sudden unity. Critics, including AI researcher Gary Marcus, argued that Amodei's proposal "would likely only wind up serving Anthropic and OpenAI" what Marcus called "regulatory capture in action." The argument is that slowing the frontier favors incumbents who already have deployed models, safety infrastructure, and regulatory relationships, while locking out newcomers.

Bernie Sanders, who separately proposed banning artificial superintelligence, responded that Amodei's call was "not enough" and that stronger government action is needed.

Amodei acknowledged the skepticism in his post: "My desire to achieve these benefits is undimmed. But the benefits will only be achieved if we build the technology in the right way."

What Happens Next

The immediate outcome is a test of whether the stated unity translates into action. Anthropic has committed to embedded evaluators. OpenAI says it will follow. Musk's xAI has not detailed specific commitments.

The deeper question is whether any voluntary industry-led slowdown can succeed where decades of nuclear and biotech safety efforts have struggled: aligning competitive incentives with collective safety. Amodei's proposal notably avoids calling for a moratorium or pause. It asks for a deliberate pace, not a stop.

As he put it, "Progress will still seem fast, and we must make wise use of the time we gain."

Sources

• Dario Amodei, "We Must Pace the Frontier" https://darioamodei.com (via Google News proxy)
• Anthony Ha, "Anthropic CEO outlines plan to slow AI development" TechCrunch, Sep 12, 2026. https://techcrunch.com/2026/09/12/anthropic-ceo-outlines-plan-to-pace-the-frontier/
• "Anthropic C.E.O. Dario Amodei Calls for A.I. Slowdown" The New York Times, Sep 13, 2026 (via Google News proxy)
• "Top AI leaders unite to warn the technology is advancing too fast" The Washington Post, Sep 13, 2026 (via Google News proxy)
• "Anthropic boss Dario Amodei calls for AI development to slow down" BBC News, Sep 12, 2026 (via Google News proxy)
• "AI's Code-Red Moment" The Atlantic, Sep 12, 2026 (via Google News proxy)
• "Rivals Altman and Musk rally behind Dario Amodei's call for an AI slowdown" Financial Times, Sep 12, 2026 (via Google News proxy)
• "Dario Amodei says OpenAI's Hugging Face hack helped convince him AI needs to slow down" Business Insider, Sep 12, 2026 (via Google News proxy)