Researchers showed that the Claude AI model could generate a working exploit for a FreeBSD kernel vulnerability (CVE-2026-4747), producing a remote root shell by building a stack overflow and ROP chain with minimal human guidance.
Posts in category: AI Security Research