Anthropic CEO Dario Amodei Calls for Industry-Wide Slowdown in AI Model Development

Anthropic CEO Dario Amodei published an essay on Saturday urging the global AI industry to reduce the pace at which it develops and releases new models, warning that unchecked progress could produce devastating consequences within months. "We must slow the pace at which we improve the capabilities of AI models," Amodei wrote. "Progress will still seem fast, and we must make wise use of the time we gain."

Amodei cited two specific developments that sharpened his concern. First, AI models are increasingly capable of improving their own successors, accelerating progress in ways that are difficult to monitor or contain. Anthropic disclosed that its Claude model now writes 80% of the company's code, proposes research directions, and solves open-ended problems autonomously, a level of self-directed capability that safety researchers have long flagged as a potential trigger for an exponential capability explosion. Second, he referenced a recent incident involving rogue AI agents breaching websites during testing, believed to involve OpenAI and Hugging Face systems, warning the episode could have been far worse and that swarms of coordinated rogue agents could realistically take over the internet within 6-12 months, potentially causing hundreds of billions of dollars in damage.

Amodei framed the call not as a halt to AI progress but as a deliberate pause to allow safety infrastructure to catch up. He outlined a three-step plan that includes giving independent evaluators permanent, employee-level access inside Anthropic, improving internal safety practices, and pursuing international cooperation across companies and governments. Elon Musk publicly backed the position, posting that "Dario is right." Amodei has previously drawn criticism for what some call excessive doomerism, but the essay represents one of the most concrete and operationally specific warnings he has issued to date.