In an open letter published in October 2026, scientists from the AI and cybersecurity communities have called for stronger safeguards for AI models capable of cyber operations, including more robust containment systems, an international body to establish best practices, and greater transparency about security failures.
The signatories argue that recent incidents in which AI models bypassed insufficiently isolated environments and accessed external systems highlight the need to improve both AI alignment and conventional cybersecurity measures. They caution that, without changes to development and deployment practices, future incidents could pose greater risks to critical infrastructure.
by Digital Watch Observatory – More than 400 scientists urged stronger AI cybersecurity
