Microsoft on June 2 unveiled MDASH (Microsoft Security Multimodal Agentic Scanning Harness), a platform that uses AI agents to automatically detect software vulnerabilities, Yahoo Finance reported on June 2.

MDASH combines more than 100 configurable AI models to search for bugs that hackers could exploit, the report said. It supports using a range of models, from the best-performing to cost-efficient ones, to avoid relying on a single model. It is currently available in an expanded preview version.

Vasu Jakkal (바수 자칼), a vice president in Microsoft’s security division, said, "The time from vulnerability discovery to exploitation is rapidly shrinking." She added, "If we do not build technology that provides context to defenders, cyber warfare will continue to flow in an asymmetric structure."

The MDASH unveiling comes as concerns spread about security attacks that misuse advanced AI models such as Anthropic’s Claude Mythos.

Mythos, unveiled in April, was not developed for cybersecurity purposes but is drawing strong interest in the sector for quickly finding vulnerabilities in major web browsers and operating systems.

Anthropic initially provided access to Mythos to about 40 companies, including Microsoft, and said on June 2 it added 150 more organisations.