Breaking

Anthropic CEO Dario Amodei Calls for Slower Frontier AI Development as Agent-Safety Incidents Mount

September 14, 2026 at 11:50 EDT

AI Agents
Security
Policy & Regulation

Anthropic CEO Dario Amodei has urged frontier AI companies to slow capability gains long enough for safety work, independent evaluation and international rules to catch up. The proposal directly complicates the market story that AI infrastructure spending can keep accelerating without a governance constraint.

September 2026 · Anthropic

Pacing the AI frontier

Dario Amodei is asking frontier labs to slow capability gains long enough for safeguards, audits and international rules to catch up.

steps in the pacing plan

6-12

months in Amodei's risk window

141,006

Claude evaluation runs reviewed

20%

monitoring-compute overhead cited by OpenAI

How large the agent coordination became

1,200

agents on message board

700

agents in the attack

Each block represents about 100 agents.

The pacing sequence

1. Embedded evaluators

Outside teams get employee-like access.

→

2. Democratic coordination

Labs align safety standards and limits.

→

3. Global coordination

Governments seek verifiable rules.

Why supporters care

Independent access could make safety claims more verifiable before models scale further.

Why critics worry

Voluntary rules and industry coordination could favor incumbent labs unless oversight is enforceable.

Continue reading

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