Breaking

AI Leaders Back Calls to Pace Frontier Development as Trump and China Resist

September 15, 2026 at 10:52 EDT

AI Agents
Policy & Regulation
Security

Anthropic CEO Dario Amodei has urged frontier AI developers to slow capability gains so safety work can catch up with systems that are starting to help build their own successors. The call has drawn support from other AI leaders but immediate resistance from President Donald Trump and from Beijing, exposing the geopolitical fault line in any attempt to pace advanced AI.

September 2026 · Frontier AI

AI leaders ask to pace the frontier

A rare push to slow advanced model development now runs into U.S.-China competition, Washington's reluctance to regulate, and evidence that AI agents can already strain real-world safeguards.

steps in Amodei's pacing plan

6-12

months in risk window

1,386

frontier AI employees signed a pacing statement

141,006

Anthropic evaluation runs reviewed

AI is accelerating AI work

Code output per engineer per day rose from a 2024 baseline to 8x in Q2 2026.

1x2024 baseline

8xQ2 2026

The proposed pacing loop

Embedded evaluators

Outside reviewers get employee-like access.

→

Shared standards

Frontier labs coordinate safety thresholds.

→

Global deal

Governments try to verify pacing across borders.

Why supporters see value

More time could improve alignment, monitoring, sandboxing and public oversight before agents reach higher-risk capabilities.

Why critics push back

Voluntary commitments may lack measurable limits, and geopolitical rivals may not accept the same speed controls.

The decision point

Loose consensus now needs enforcement

The central question is whether embedded evaluators and shared thresholds become verifiable constraints, or whether commercial pressure and national competition keep frontier development moving faster than oversight.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.