Breaking

Anthropic CEO Dario Amodei Urges Frontier AI Slowdown as OpenAI, DeepMind and Musk Signal Support

September 14, 2026 at 23:51 EDT

Foundation Models
AI Agents
Policy & Regulation

Anthropic CEO Dario Amodei has called for frontier AI developers to slow capability gains so safety work can catch up, turning a long-running AI-risk debate into an immediate industry flashpoint. His September essay, We Must Pace the Frontier, proposes embedded outside evaluators, coordinated safety standards among democratic countries, and eventual global coordination, while stressing that pacing does not mean stopping model training.

September 2026 · Frontier AI

AI labs debate slowing the frontier

Anthropic's Dario Amodei called for pacing advanced AI development as OpenAI, Google DeepMind and Elon Musk signaled support. The debate turns on whether safety oversight can keep up with increasingly autonomous agents.

steps in Amodei's pacing plan

6-12

months in his internet-risk warning

80%+

of merged Anthropic code authored by Claude

Why the pace question escalated

1x2024 code baseline

8xQ2 2026 output

Anthropic says AI-assisted development sharply accelerated internal code output, while cautioning that lines of code are an imperfect productivity measure.

Supportive case

Independent evaluators and shared safety bars could make lab commitments more verifiable before more capable agents arrive.

Critical case

Skeptics warn that coordination among leading labs could entrench incumbents and substitute voluntary oversight for public regulation.

The proposed safety loop

Embed evaluators→Verify incidents→Set shared bars→Pace capability gains

Continue reading

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