AI Just Hacked One Of The World's Most Secure Operating Systems

TL;DR AI
2 min readKey summary
An autonomous AI agent developed two working exploits for FreeBSD's CVE-2026-4747 kernel flaw.
The agent produced a root shell on unpatched systems after about four hours of compute time.
The vulnerability is in the RPCSEC_GSS NFS module and allows a stack buffer overflow without authentication.
The exploit chain required multi-packet shellcode, kernel thread hijacking, and several kernel-to-user transitions.


