Microsoft launched MDASH, an AI-driven security platform to automate large-scale code auditing for Windows and Azure. The system utilizes over 100 specialized AI agents to identify, validate, and prove security flaws in complex codebases. This multi-agent approach shifts cybersecurity away from single AI models toward integrated orchestration.
The MDASH platform remains model-agnostic to allow the integration of various AI models into its validation and proof generation workflow. Microsoft reported the system achieved a top score on the public CyberGym benchmark. This performance outperformed the next leading entry in the security competition.
This development reflects Microsoft's increasing focus on using complex AI systems to secure its core products.