AI Agents Got Root on Linux via 948 Kernel Flaws in OpenAI Research

An OpenAI cyber model was able to find and exploit a large number of Linux kernel privilege escalation bugs, totaling 948 distinct flaws. This was achieved for a cost of approximately $1.2 million. The model was able to successfully gain root access on Linux systems by exploiting these vulnerabilities. The research was conducted by OpenAI and the findings have been reported. The model's ability to identify and exploit such a large number of kernel flaws is notable. The research highlights the potential for AI-powered tools to discover and exploit vulnerabilities in Linux kernels.