Defend at AI Speed: Microsoft’s New Multi-Model Agent Security System Exceeds Leading Industry Benchmarks | Microsoft Security Blog

Today, Microsoft announced a major advance in AI-powered cyber defense: Our new agent security system helped researchers find 16 new vulnerabilities in the Windows networking and authentication stack – including four critical remote code execution flaws in components such as the Windows kernel TCP/IP stack and the IKEv2 service. They used the new Microsoft security MUlti-MonDel Agenetic SCan Harness (codename MDASH), created by Microsoft’s Autonomous Code Security team. Unlike single-model approaches, the system orchestrates more than 100 specialized AI agents across an ensemble of boundary models and distilled models to consistently discover, discuss, and prove exploitable bugs.

The results speak for themselves: 21 of 21 hidden vulnerabilities were found in a private test driver without any false alarms; 96% recall against five years of confirmed Microsoft Security Response Center (MSRC) cases in clfs.sys and 100% in…