Breaking

Anthropic CEO Dario Amodei Urges Slower, More Auditable AI Development

September 13, 2026 at 11:05 EDT

Foundation Models
AI Agents
Policy & Regulation

Anthropic CEO Dario Amodei is calling for frontier AI developers to slow the pace of capability gains so safety work can catch up. He is proposing permanent outside oversight and coordination among rival labs, while OpenAI CEO Sam Altman has said OpenAI will adopt the evaluator measure. Amodei’s proposal

September 12, 2026 · Anthropic

A slower AI frontier, with safety in the room

Dario Amodei says capability gains should be paced so alignment, testing and independent oversight can keep up.

The proposal in numbers

steps in the pacing plan

1-2

years Amodei says pacing could buy for safety work

481M

Anthropic transcripts screened

The warning window

Amodei warns that a similar agent swarm could become capable of taking over the internet within this range.

6 months

12 months

Three steps to pace the frontier

Step 1Embedded evaluators

Independent reviewers inspect safety work from inside the company.

→

Step 2Democratic coordination

Frontier companies align on standards and limits for unchecked progress.

→

Step 3Global coordination

Governments seek workable agreements on the most dangerous uses.

Why the proposal draws both support and concern

Potential upside

More time for alignment and interpretability
Independent checks on internal claims
Greater visibility into incidents

Open questions

Can rivals coordinate without antitrust barriers?
Can global compliance be verified?
Could rules favor the largest labs?

Continue reading

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