OpenAI released ChatGPT Cyber, a model built for security work, to researchers in Daybreak Red, the top tier of its program that gives vetted cyber defenders access to frontier models, a day after saying the model flagged 948 potential issues in the Linux kernel and prepared patches for them. Charge Robotics CTO Max Justicz told TechCrunch that GPT 5.6 Sol, which he used through the program’s Daybreak Blue tier, took about two hours to find a flaw that could let attackers take over Ubuntu servers. Debian has since patched the flaw, and there’s no known evidence anyone exploited it.
