Amodei's new essay calls for a deliberate slowdown in AI development
Anthropic CEO Dario Amodei calls for embedded evaluators and international coordination. Altman, Hassabis, and Musk publicly back the call.
Illustration · AI-generated (AI IN LIFE)
At a glance
- Published: Saturday, September 12, 2026, by Anthropic CEO Dario Amodei
- Essay title: The Adolescence of Technology, roughly 3,800 words
- Publicly cosigned by Sam Altman (OpenAI), Demis Hassabis (Google DeepMind), and Elon Musk (xAI)
- OpenAI, Anthropic, and Google DeepMind confirmed weeks of joint safety talks
- Three-stage plan: embedded evaluators, coordination among democracies, international agreements
Anthropic CEO Dario Amodei published a roughly 3,800-word essay titled The Adolescence of Technology on Saturday, September 12, 2026. In it, he calls for a deliberate, coordinated slowdown of AI development, arguing that self-improving systems are now advancing faster than researchers can understand or secure them.
His three-stage proposal starts with independent evaluators embedded directly inside the leading labs, moves to coordination among AI companies in democratic countries, and aims longer-term at international agreements. He argues that recursive self-improvement, where an AI system helps design its own successor, must be pursued very carefully, if at all.
The call drew unusually broad support: OpenAI chief Sam Altman, Google DeepMind chief Demis Hassabis, and xAI founder Elon Musk all publicly cosigned it. Days later, OpenAI policy chief Chris Lehane confirmed that OpenAI, Anthropic, and Google DeepMind had already been working together on safety measures for weeks, without needing an antitrust waiver to do so.
Critics note that a voluntary commitment without legal force carries little enforcement power, and that it is precisely the companies benefiting most from the current pace that are now publicly urging caution.
FAQ
What is recursive self-improvement?
It means an AI system actively helps design or train its own successor. Amodei warns this process could make development harder to control if it moves too fast.
Is this a binding development moratorium?
No. Amodei isn't demanding an immediate halt, he wants embedded evaluators and coordinated rules among the leading labs — a voluntary commitment, not a legal requirement.
Why is the broad support notable?
Because Amodei, Altman, Hassabis, and Musk are normally fierce competitors. Their joint public backing is considered the real news here, more than the essay itself.
