Microsoft AI Finds 59 Linux Kernel Bugs, With Root Exploits Demonstrated for Eight
Attack Tradecraft · 4 sources · Updated 3d ago · First seen Nov 12, 2026

Microsoft’s FORGE Lab reported 59 Linux kernel security bugs discovered using its AI-driven MDASH scanning system, resulting in 33 assigned CVEs and 38 fixes merged into the mainline kernel. Researchers demonstrated local root privilege escalation for eight vulnerabilities in laboratory conditions, although some demonstrations required non-default configurations or disabled protections. The bugs had existed for an average of 8.1 years, with the oldest dating back at least 21 years. One highlighted flaw, CVE-2026-64563, is a use-after-free in the kernel’s rhashtable implementation that can be exploited through unprivileged Netlink socket diagnostics to escalate privileges on Linux and Android.
The findings underscore a growing gap between AI-assisted vulnerability discovery and the human capacity to validate reports, review patches, and coordinate remediation. Linux maintainer Lorenzo Stoakes withdrew from several reviewing responsibilities, while stable maintainer Greg Kroah-Hartman warned of reviewer shortages and maintainer burnout. In the Linux 7.4-rc2 announcement, Linus Torvalds noted tentative progress in the Berkeley Packet Filter subsystem, where fixes reportedly began outpacing new reports, but cautioned that other areas had not necessarily reached that point. Security teams should track distribution and Android vendor advisories, confirm whether relevant fixes have reached deployed kernels, and prioritize affected systems where untrusted users or workloads can execute locally; mainline fixes alone do not establish that downstream systems are patched.


