Microsoft Introduces MDASH for Large-Scale AI Vulnerability Research

Microsoft has unveiled MDASH, a multi-model agentic security platform using over 100 specialized AI agents to automate large-scale vulnerability discovery across Windows, Hyper-V, Azure, and other Microsoft codebases. The system achieved 88.45% on the CyberGym benchmark, outperforming competitors by ~5 points, and reported 96–100% recall on internal historical vulnerability datasets. MDASH operates as a multi-stage pipeline where agents handle scanning, debate, validation, deduplication, and exploitation separately. Microsoft emphasizes the orchestration framework matters more than any single model, and the system is model-agnostic by design. It is currently in limited private preview, with discussion emerging around governance risks of large-scale agentic security systems.