Microsoft Used AI to Hunt Linux Kernel Bugs: 26,502 Findings Became 59 Real Reports
Microsoft's FORGE Lab used an internal system, codenamed MDASH, to scan selected Linux kernel areas such as networking, filesystems and security modules from late May to mid-June 2026. One full-time researcher and one intern led the three-month effort. So far it has produced 59 public upstream reports and 33 assigned CVEs. The team confirmed working root privilege-escalation exploits for eight of those CVEs, and fixes for 38 reports have been merged into the mainline kernel.
The initial scans produced 26,502 raw findings in about two weeks. Low-confidence claims were filtered out using the system's confidence levels and separate screening agents, leaving 11,769 candidates. AI agents then tried to reproduce each issue in a QEMU test environment, and experts reviewed the 621 cases that came with proof-of-concept evidence. They identified root causes, merged duplicates and prepared submissions for Linux maintainers. In the end, 59 reports came from 621 cases (9.5%).
The researchers say finding the bugs was only part of the work. Turning results into reproducible bugs, reports and patches took months of expert investigation and collaboration with kernel maintainers. For defenders, the lesson is that AI can speed up vulnerability discovery, but human validation remains essential. It also suggests a steady flow of kernel fixes, so keeping Linux systems patched matters.