Today, as artificial intelligence technology is developing rapidly, Microsoft once again has shown its innovative power to us. On May 13th, Microsoft's own code security team officially released a multi-model agent scanning framework called MDASH. This new security system not only leads in technology but also demonstrates remarkable vulnerability identification capabilities in practical applications, especially standing out when compared with the well-known GPT-5.5 and Anthropic's Mythos models.

The uniqueness of MDASH lies in the fact that it does not rely on a single AI model, but instead adopts a multi-agent collaboration strategy. Microsoft's team integrated over 100 different AI agents in this framework, each responsible for different tasks, covering multiple stages from code preparation to vulnerability scanning and result verification. This way, MDASH can use the most powerful model for core reasoning tasks, while for scenarios requiring rapid processing of massive code, it leverages lightweight models that respond quickly.

In the latest CyberGym public benchmark test, MDASH not only successfully identified 16 previously unknown vulnerabilities, including four high-risk remote code execution vulnerabilities, but also achieved a perfect 100% identification rate in private tests facing 21 deliberately implanted vulnerabilities, with zero false positives throughout the process. This achievement undoubtedly provides strong support for Microsoft's internal security protection work.

Additionally, MDASH also performed exceptionally well in retrospective testing of historical vulnerabilities, such as achieving a 96% recovery rate for clfs.sys's vulnerabilities over the past five years, and a 100% recovery rate for tcpip.sys. Such high efficiency and accuracy have greatly boosted Microsoft's confidence in product security enhancement.

Currently, MDASH has started assisting Microsoft's internal engineering teams in product security enhancement, and is also offering internal preview testing to limited customers. These measures mark another important breakthrough for Microsoft in the field of software security, and in the future, MDASH will undoubtedly become an indispensable tool in the field of security protection.

Key Points:

🔍 MDASH is Microsoft's newly released multi-model agent scanning framework, surpassing the GPT-5.5 and Mythos models.

⚡ The system uses over 100 AI agents working together, accurately covering stages such as code preparation and vulnerability scanning.