Breaking

Anthropic’s Dario Amodei Urges a Slower AI Frontier as Rivals Back Safety Push

September 13, 2026 at 06:15 EDT

Foundation Models
Policy & Regulation
AI Agents

Anthropic CEO Dario Amodei has called for the frontier AI industry to slow the pace at which model capabilities improve, arguing that safety and oversight are falling behind. OpenAI CEO Sam Altman and xAI chief Elon Musk backed the call, while Google DeepMind chair Demis Hassabis endorsed its direction but said the details still need work. Amodei’s proposal

SEPTEMBER 2026 · ANTHROPIC

Frontier AI leaders call for a slower pace

Amodei’s plan asks competitors to give safety and independent verification time to catch up. The concrete commitments remain mostly unilateral.

The proposal in numbers

6–12

months in Amodei’s risk window

1,386

employees signed the pacing statement

weeks of OpenAI reinforcement-learning pause

steps in the pacing framework

Risk window cited by Amodei

The warning describes a possible escalation within a 6–12 month window.

6 mo

lower bound

12 mo

upper bound

The upper end of the cited window is twice the lower end.

How pacing is meant to work

Embedded evaluators

Independent reviewers verify safety practices and report incidents.

→

Shared standards

Democratic countries coordinate safety standards and limits.

→

Global coordination

Governments work toward wider agreements and verification.

Why the consensus is incomplete

What supporters see

More time for alignment
Independent verification
Progress remains possible

What remains unresolved

No common cap yet
Antitrust questions
Global verification is hard

Continue reading

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