Microsoft has introduced MDASH (Multi-Model Agentic Scanning Harness), a security solution that uses over 100 specialized AI agents to identify software vulnerabilities. On May 12, 2026, MDASH identified 16 new vulnerabilities (CVEs) in the Windows networking and authentication stack, four of which were critical, including remote code execution vulnerabilities in tcpip.sys, ikeext.dll, netlogon.dll, and dnsapi.dll. Ten of these vulnerabilities can be accessed over the network without authentication.
MDASH operates through a four-stage pipeline: analyzing source code, scrutinizing for suspicious elements, debating the exploitability of issues, and attempting to exploit vulnerabilities. The system is model-agnostic and allows integration of new models and domain-specific knowledge.
MDASH scored 88.45 percent on the CyberGym benchmark, ranking first among competitors, although the comparison may not be entirely fair as it contrasts a comprehensive framework with individual models. The models used to achieve this score are not specified. MDASH is supported by Microsoft's Autonomous Code Security Team and is currently in a limited private preview for select customers.