Microsoft has announced MDASH (Multi-moDel Agentic Scanning Harness), an artificial intelligence (AI) agent system that can detect security vulnerabilities like Claude Mythos. MDASH uses over 100 AI agents working together using multiple models.
MDSAH has already found 16 vulnerabilities in Windows that were fixed in this week’s Patch Tuesday update. It also found all 21 vulnerabilities intentionally placed in software in its tests. In the CyberGym security benchmark test, it scored 88.45%, which is the highest and bested Claude Mythos.
AI that can detect security vulnerabilities is a current trend, with OpenAI also announcing their system, Daybreak, this week.
MDASH is currently being used by Microsoft’s engineering team to significantly improve security systems using existing AI models. It is also being tested by several customers as part of a limited test.
