Breaking

Anthropic CEO Calls for Slower Frontier AI as OpenAI Backs Independent Evaluators

September 12, 2026 at 14:00 EDT

Policy & Regulation
Security
AI Agents

Anthropic CEO Dario Amodei is calling on frontier AI developers to slow the rate of capability gains, arguing that alignment and security work need time to catch up. OpenAI CEO Sam Altman has backed pacing and said OpenAI will make the same independent-evaluator commitment.

September 12, 2026 · Anthropic

A Slower, More Verifiable AI Frontier

Dario Amodei is calling for slower capability gains and embedded independent evaluators. Sam Altman says OpenAI will make the same evaluator commitment.

The measurable signal

Code output accelerated

Anthropic reports 8× as much code merged per engineer day in Q2 2026 as in 2024, while cautioning that line count is an imperfect productivity proxy.

1×

2024 baseline

8×

Q2 2026

Key figures

80%+

of merged code authored by Claude by May 2026

6–12

months in Amodei's projected risk window

steps in the pacing framework

Core distinction

Pacing is not a halt

Model training and technical progress continue. The aim is to create time for alignment, interpretability, testing, and operational safeguards.

The central test: can independent evaluators verify what companies claim?

Potential upside

More time, more accountability

More time for alignment, interpretability, testing, and operational excellence.

Outside reviewers can check practices and report incidents.

Open questions

Enforcement remains difficult

Industry coordination may need antitrust waivers.

Global compliance must be verifiable; critics warn of protectionist incentives.

Continue reading

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