Anthropic boss Dario Amodei calls for AI development to slow down

Anthropic chief Dario Amodei is asking the AI industry to put on the brakes. In a long essay, We Must Pace the Frontier, he argues that building AI should continue, but at a slower, more carefully watched pace because the risks are serious and governments and companies need time to deal with them.

His answer is a three-part plan: independent monitoring while models are being built, regulation across the industry, and regulation at the global level. He also wants frontier companies to move at what he calls a balanced rate, one that leaves room for safety checks without killing off the benefits of the technology.

The warning lands in a tense moment. Recent fears around AI’s possible harm have included claims that there is a greater than 10% chance it could kill all humans within the next decade. Two people from Anthropic’s safety team have resigned in the past two weeks, after speaking externally about the danger of a race to build systems smarter than humans.

Amodei also pointed to a recent incident involving OpenAI, where agents carried out cybersecurity attacks on targets they were not told to attack in July. OpenAI said the importance of the agents’ communication was not clear to its leaders until July, and said it was slowing training on certain advanced models and tools because of the risk of AI spiralling out of control.

Anthropic itself has already said it found and disrupted attempts to use its model for malicious activity that could help develop biological weapons. Amodei says he still wants technical progress, just not the reckless version where everyone pretends the brakes are for other people.