en.Wedoany.com Reported - Microsoft has released the MDASH system, which uses over 100 AI agents for vulnerability detection, in response to the rise of AI programs tracking security flaws. Microsoft stated that MDASH discovered 16 new Windows-related vulnerabilities, including four critical remote code execution flaws in components such as the Windows kernel TCP/IP stack and the IKEv2 service. MDASH achieved a score of 88.45% on the CyberGym benchmark, which evaluates AI agents' ability to discover software defects, surpassing AI models such as Anthropic's Claude Mythos and OpenAI's GPT 5.5.

Microsoft noted in its announcement: "AI vulnerability discovery has shifted from a research interest to enterprise-grade, production-level defense." The company stated that its system demonstrates a "persistent advantage" by leveraging multiple AI models rather than relying on a single model. Microsoft developed over 100 AI agents, each using cutting-edge AI models and efficient small models, specifically for vulnerability detection. Microsoft added: "No single model is optimal at all stages. The multi-model agent scanner runs a configurable panel of models." After AI agents scan computer code to find vulnerabilities, they engage in debate to confirm whether the findings are consistent. Microsoft stated: "Disagreement between models is itself a signal: when a reviewer flags something as suspicious and a debater cannot refute it, the posterior credibility of the finding increases." MDASH is already being used by Microsoft's security engineering team and a small group of customers in a limited private preview to prevent misuse, as it "can approach the level of professional attack researchers." Microsoft is opening access to enterprise customers who apply.

Against the backdrop of hackers using AI models to find software flaws or orchestrate attacks, the cybersecurity industry has entered an arms race: AI tools have the potential to strengthen defenses, but they can equally fall into malicious hands and cause damage. The application of MDASH demonstrates that Microsoft, through AI agents for vulnerability detection, is attempting to verify whether AI can fortify software systems to withstand AI-driven attacks.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com