Microsoft’s new multi-model agentic security system tops industry benchmark
Microsoft announced its new multi-model agentic security system (MDASH) has achieved top performance in an industry benchmark. The system, which coordinates over 100 specialized AI agents across frontier and distilled models, discovered 16 new vulnerabilities in the Windows networking and authentication stack, including four critical remote code execution flaws.
Article intelligence
Key points
- Microsoft's MDASH system uses over 100 AI agents to discover 16 new Windows vulnerabilities
- Includes four critical remote code execution flaws in kernel TCP/IP stack and IKEv2 service
- The system tops industry benchmark, marking a milestone in AI-powered cyber defense
Why it matters
This matters because microsoft's MDASH system uses over 100 AI agents to discover 16 new Windows vulnerabilities.
Technical impact
May affect model selection, inference cost, product capability, and evaluation benchmarks.
Today Microsoft announced a major step forward in AI-powered cyber defense: our new agentic security system helped researchers find 16 new vulnerabilities across 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-model agentic scanning harness (codename MDASH) which was built by Microsoft’s Autonomous Code Security team. Unlike single-model approaches, the harness orchestrates more than 100 specialized AI agents across an ensemble of frontier and distilled models to discover, debate, and prove exploitable bugs end-to-end.
Learn more and sign up to join the preview