Breaking

Anthropic's Dario Amodei Calls for Slower Frontier AI Development After Agent Incidents

September 14, 2026 at 11:29 EDT

Policy & Regulation
AI Agents
Foundation Models

Anthropic CEO Dario Amodei has called for frontier AI labs to slow the rate of capability gains, arguing that safety systems, outside audits and international governance are falling behind the technology. The proposal commits Anthropic to embedded third-party evaluators and has quickly become a test of whether leading AI companies can make voluntary restraint verifiable.

September 2026 | Anthropic

Frontier AI Gets a Speed-Limit Proposal

Dario Amodei is asking labs to slow capability gains enough for safeguards, independent review and global coordination to catch up.

pacing stages

1,386

employees backed pacing tools

70,000+

agent messages and files

code shipped per engineer

The Measured Acceleration

At Anthropic, engineers now ship eight times as much code per quarter as they did from 2021 to 2025.

2021-2025 baseline

current rate

How the Pacing Plan Scales

1. Embedded evaluators

Outside teams get ongoing access to verify safety practices.

2. Democratic coordination

Labs and governments align on common frontier standards.

3. Global coordination

Countries negotiate limits on the highest-risk capabilities.

Why backers like it

Auditors would see more than polished release documents.
Extra time could improve alignment, security and interpretability.

What critics fear

Company-selected monitors may not be enough.
Shared limits could favor large closed-model labs.

Continue reading

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