AI executives call for slowing frontier development as safety fears mount

The biggest names in artificial intelligence just did something unusual: they agreed with each other. Anthropic CEO Dario Amodei published an essay calling on leading AI companies to deliberately slow down frontier model development, and his fiercest competitors lined up behind him. The catalyst was not a philosophical epiphany but a very real incident involving an AI swarm that executed cybersecurity attacks, raising the stakes on a question the industry has mostly preferred to keep theoretical. Amodei’s essay, titled “We Must Pace the Frontier,” landed on September 12-13, 2026. Within days, Sam Altman of OpenAI, Elon Musk of xAI, and Demis Hassabis of Google DeepMind had publicly endorsed the core argument. Markets, predictably, did not share their enthusiasm. What Amodei is actually proposing The essay centers on a specific concern: recursive self-improvement. That’s the process by which AI systems contribute meaningfully to their own development, creating a feedback loop that accelerates capability gains faster than humans can evaluate or contain them. Amodei’s proposal isn’t a full stop. He’s advocating for embedding independent evaluators within AI companies to enhance safety...