Microsoft just dropped what might be the most consequential AI security tool of the year, and it has nothing to do with chatbots writing your emails. The company’s new MDASH system, short for Microsoft Security multi-model agentic scanning harness, uses more than 100 specialized AI agents working together to find vulnerabilities in code before attackers do.

In its debut, the system discovered 16 previously unknown vulnerabilities in Windows networking and authentication components. Four of those were critical remote code execution flaws. In English: the kind of bugs that let someone take over your machine from across the internet.

How MDASH actually works

Think of MDASH less like a single security scanner and more like an entire security team compressed into software. The system runs a multi-stage pipeline covering preparation, scanning, validation, deduplication, and proofing. Each stage uses different specialized AI models rather than relying on a single large language model to do everything.

The results speak for themselves. MDASH scored 88.45% on the public CyberGym benchmark, which tests real-world vulnerability detection capabilities. That’s roughly five points ahead of the nearest competitor. Internal testing was even more impressive: a 96% recall rate on historical clfs.sys cases and a perfect 100% on tcpip.sys tests. Zero false positives on 21 planted vulnerabilities.

Microsoft also released a companion model called MAI-Cyber-1-Flash, which scored 96% on the CyberGym benchmark as a standalone system.

All 16 newly discovered vulnerabilities were patched in the May 2026 Patch Tuesday update. So the system isn’t just finding bugs in a lab. It’s finding bugs that ship in fixes to billions of devices.

Why crypto and Web3 should care

MDASH supports all major programming languages and can accommodate proprietary codebases. It features an extensible pipeline with domain-specific plugins. That architecture means it could, in theory, be adapted for Solidity, Rust, or Move-based smart contract auditing.

The system is currently available in a limited private preview through Azure commercial cloud regions. Key partners like CyberOne are already engaged.

The competitive landscape shifts

Microsoft’s move puts additional pressure on dedicated cybersecurity companies. The 88.45% CyberGym score establishes a new benchmark that competitors will need to match or explain away.

Microsoft isn’t publishing a research paper and calling it a day. They found real bugs, shipped real patches, and are onboarding real customers through the private preview phase.