Microsoft Intros New Agentic AI Security Multi-Model Defense System
AI Summary: Microsoft has introduced a new multi-model agentic AI security system, codenamed MDASH, developed by its Autonomous Code Security team. This system successfully identified 16 previously unknown vulnerabilities in the Windows networking and authentication stack, including four critical remote code execution flaws. MDASH operates with over 100 specialized AI agents, achieving an 88.45 percent score on the CyberGym benchmark, which evaluates real-world vulnerabilities. The initiative reflects Microsoft's strategy to integrate AI as a fundamental component of cybersecurity operations, enhancing threat detection and remediation processes.