A specialized OpenAI model found a datacenter-wide flaw in Canonical software
From the AI digest for Nov 5, 2026
What happened: Max Justicz, chief technology officer of Charge Robotics, says he used a specialized OpenAI model to discover a vulnerability in popular open-source software maintained by Canonical, the British company behind Ubuntu — a flaw he says would let someone hack into major datacenters. Justicz says he alerted Canonical.
What's genuinely new: This is a named person, a named software maintainer and a concrete disclosed vulnerability — the AI-led vulnerability-discovery story moving from OpenAI's own pipeline-run numbers to an independent researcher's real-world find against infrastructure that datacenters rely on. Canonical's response and the specific component are not in the reporting.
Watch: Canonical's confirmation, patch or advisory; which software component and what severity; and whether the "specialized OpenAI model" is the unreleased cyber model OpenAI's own November pipeline run pointed to.
Links: The Information · Nov 4, 7:00 AM ET