Breaking

Anthropic CEO Dario Amodei Urges AI Labs to Slow Frontier Development for Safety

September 14, 2026 at 23:43 EDT

Foundation Models
AI Agents
Policy & Regulation

Anthropic CEO Dario Amodei has called on leading AI labs to slow the pace of frontier-model development, arguing that safety work and independent oversight are falling behind the speed of capability gains. In a September 2026 essay, Amodei proposed what he called "pacing the frontier": a coordinated effort to give companies and outside evaluators more time to align, test and safeguard advanced AI systems before the next major capability jumps.

September 2026 · Anthropic

A call to pace frontier AI

Dario Amodei is urging leading AI labs to slow capability gains long enough for safety work, outside evaluation and international coordination to catch up.

parts in the pacing framework

6-12

months in one risk window cited

3,800

words in the reported essay length

The proposed safety stack

3 steps

A framework, not a hard speed limit

The proposal layers embedded evaluators, shared democratic standards and wider global coordination instead of defining a single technical cap.

1 · Embedded review

Outside evaluators get broad access inside labs.

2 · Shared standards

Democratic governments and companies align rules.

3 · Global pacing

Coordination would need to include geopolitical rivals.

Why supporters back it

Broad evaluator access could make safety claims more testable before powerful systems are released.

Why critics object

Voluntary pacing may be hard to enforce and could favor the biggest labs over smaller competitors.

Continue reading

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