Anthropic CEO says it’s time to pump the brakes on AI (2026-9-13)

https://www.theverge.com/ai-artificial-intelligence/994337/anthropic-ceo-slow-down-ai-development

Anthropic CEO Dario Amodei proposed a three-step plan to slow down frontier artificial intelligence development, following earlier warnings from co-founder Jack Clark about needing an AI "brake pedal", as detailed in The Verge report. [1, 2]

The Three-Step Slowdown Plan

Independent Evaluators: AI companies should allow deep, employee-like access for independent safety experts.
Industry Standards: Competitors must coordinate to set shared safety limits.
International Rules: Governments need to cooperate so companies are not punished competitively for slowing down. [1, 2]

The "Brake Pedal" Warning

Recursive Self-Improvement: Co-founder Jack Clark warned that AI models will soon improve themselves without human help.
Loss of Control: Without a brake pedal, humans risk losing control over fast-moving superintelligence.
The Nuclear Analogy: Anthropic compares the needed restraint to a nuclear nonproliferation treaty. [1, 2]

Industry and Public Reactions

Rival Support: Leaders Sam Altman of OpenAI and Elon Musk of xAI agreed with the call for deceleration and independent oversight.
Employee Concerns: Safety debates intensified after an Anthropic researcher resigned and warned about extreme risks. [1, 2]

If you'd like, I can provide:

• More details on Dario Amodei's three-step framework
• Context on Jack Clark's original comments
• Reactions from other tech leaders and lawmakers

Let me know what you want to explore next.