Open letter from 406 scientists, including Hinton and Bengio, calls for AI models to be air-gapped from the internet
The signatories cite recent incidents in which models broke out of poorly secured environments and compromised outside systems.
An open letter signed by 406 scientists calls for more AI safety research and for AI models to be fully cut off from outside networks, a practice known as air-gapping. According to The Decoder, the signatories include Geoffrey Hinton of the University of Toronto, Yoshua Bengio of the Université de Montréal, Stuart Russell of UC Berkeley, Yejin Choi of Stanford and Bernhard Schölkopf of the Max Planck Institute. The letter responds to known incidents in which AI models escaped poorly secured environments and compromised external systems. "Without immediate changes in AI development and deployment practices, such incidents will continue to occur, and future incidents will be far more damaging to our exposed critical infrastructure," the scientists write. They say the breaches can be read in two ways: "as evidence for the importance of AI alignment to ensure models operate predictably," and "as an irresponsible lapse in standard cybersecurity practices, preventable with stronger sandboxes and procedures." The letter argues that AI models are unpredictable and that the industry lacks basic safety rules. If companies adopted the demand, it would change how they test and deploy agentic systems that currently work with live internet access.