Microsoft Intros New Agentic AI Security Multi-Model Defense System
A brand-new multi-model agentic AI security system constructed by Microsoft’s Autonomous Code Security team helped scientists find 16 new vulnerabilities throughout the Windows networking and authentication stack, the company anounced in a recent security post.
The Redmond-based company is hedging its future security operations fixates the use of collaborated AI agents to help its employees carrying out traditional security operations.
Microsoft said its internal system, codenamed MDASH (Microsoft Security multi-model agentic scanning harness), helped researchers discover 16 formerly unknown vulnerabilities across Windows networking and authentication parts, consisting of 4 vital remote code execution defects.
Unlike standard AI security tools that count on a single model, the software application giant said, “MDASH collaborates more than 100 specialized AI agents running across multiple frontiers and distilled models.”
Microsoft said the system accomplished industry-leading benchmark outcomes, consisting of an 88.45 percent score on the CyberGym criteria, which covers more than 1,500 real-world vulnerabilities.
MDASH offers insight into Microsoft’s broader push toward what it calls “agentic security,” where autonomous AI systems significantly assist– and in many cases, automate– risk detection, investigation, and removal.
Scientists from Group Atlanta, the group that won $20 million in DARPA’s AI Cyber Difficulty, helped create MDASH. Microsoft is placing research study into MDASH as part of an overarching approach to turn AI-powered vulnerability research into scalable, production-grade security engineering.
Taesoo Kim, Microsoft’s vice president of agentic security, noted that the system is designed to evaluate code autonomously, argument exploitability, validate findings and produce proof-of-concept exploits.
This workout illustrated Microsoft’s positioning of AI not simply as an efficiency tool for protectors, but as a core operational layer for determining and mitigating vulnerabilities before opponents can exploit them.
For more information, checked out the Microsoft blog site.