Dario Amodei and fellow AI leaders are calling to ‘Pace the Frontier’—but what’s the plan?

Dario Amodei Outlines Plan to Slow AI Frontier Development

A week after an Anthropic researcher's grim warning about AI risks unsettled the industry, Anthropic CEO Dario Amodei has laid out his proposal to "pace the frontier" of AI development. The plan centers on independent safety evaluators and closer coordination among AI labs in democratic countries, and it has already drawn both endorsements from parts of the industry and sharp criticism from Nvidia CEO Jensen Huang.

Why Pacing the Frontier Matters

Amodei's proposal arrives at a moment of heightened anxiety within the AI community. Coming on the heels of a fellow Anthropic researcher's doomsday-style warning, the plan represents an attempt to translate abstract safety concerns into concrete policy mechanisms. The core idea is that development of the most powerful AI systems should not be a free-for-all race, but rather a measured process with checkpoints built in.

Central to the framework is the role of independent safety evaluators. Rather than allowing companies to grade their own homework, Amodei envisions third-party assessments determining whether frontier models are safe enough to proceed to the next stage of development or deployment. This would create an external check on the speed at which increasingly capable systems reach the public.

Coordination Among Democratic Nations

The second pillar of the plan calls for AI laboratories in democratic countries to align their practices. The logic is that if leading labs in the United States and allied nations move in step with one another, no single actor gains an advantage by cutting corners on safety. Coordination, in this view, removes the competitive pressure that might otherwise push labs to rush ahead.

Industry Reaction: Support and Pushback

The proposal has not landed uniformly across the industry. Some figures have voiced support for the general direction, seeing external evaluation as a sensible safeguard. Others, most prominently Nvidia's Jensen Huang, have pushed back. Huang's objections reflect a broader tension in Silicon Valley between those who favor deliberate pacing and those who argue that slowing down cedes ground to competitors, particularly in China, who may not adopt similar restraints.

The disagreement highlights a fault line that will only deepen as models grow more capable: whether safety can be enforced through coordination, or whether market and geopolitical competition will always outrun voluntary agreements.

What Comes Next

Whether Amodei's framing gains traction with policymakers remains to be seen. The coming months will show if independent evaluation and international lab coordination move from proposal to practice, and whether other frontier labs sign on.