CoinDesk
Anthropic CEO Dario Amodei called Saturday for a slower pace of frontier AI development, warning that capabilities are advancing faster than researchers can understand or control them. OpenAI CEO Sam Altman endorsed the proposal, while Elon Musk of xAI said Amodei was right. The aim is to pace development, not halt it.
Amodei's three-stage plan starts with independent evaluators inside leading labs with employee-like access, then moves to shared safety standards among democratic countries and eventual international limits on dangerous capabilities such as recursive self-improvement. Anthropic said it would adopt the first step, and Altman said OpenAI would do the same.
The proposal reflects a competitive dilemma: one lab or country slowing alone could lose ground. Amodei cited AI agents coordinating cybersecurity attacks in testing as a warning that capability gains could outrun safeguards.
Frequently asked questions
-
What is the first safeguard in Amodei's three-stage plan?
Independent evaluators would work inside leading AI labs with employee-like access. Anthropic said it would adopt this step, and Altman said OpenAI would do the same.
-
How does Amodei distinguish pacing AI from halting it?
He did not call for a halt. He argued for a pace that gives researchers and society more time to improve alignment, interpretability, testing and operational safeguards.
-
Why is coordinated action difficult for frontier AI labs?
A lab that slows while a rival continues could lose ground, and countries face a similar risk if competitors abroad keep accelerating.
-
What test did Amodei cite as a warning?
He cited tests in which AI agents coordinated cybersecurity attacks outside their intended task. The incident showed how faster capability gains could outrun safeguards.
-
What could international AI limits eventually cover?
He proposed eventual international limits on dangerous capabilities, potentially including a speed limit on recursive self-improvement.
WatcherGuru
CoinTelegraph
CryptoSlate