Microsoft launches its first cybersecurity model, plus a new agentic cybersecurity system - TechCrunch

What the report says

TechCrunch reported that Microsoft on Monday introduced its first cybersecurity-focused AI model, MAI-Cyber-1-Flash, along with a new agentic security platform called Perception during a small event in San Francisco. The company said the model is intended to help identify difficult vulnerabilities in large and complex codebases and is designed to work with MDASH, Microsoft’s system for finding and fixing software security flaws.

According to the report, Perception is built to coordinate groups of AI agents across security workflows, including bug discovery, prioritization and remediation. Microsoft described the system as using agentic red, blue and green teams: red teams simulate possible attacks, blue teams detect and triage issues, and green teams take corrective steps. The platform can also connect with MDASH.

Microsoft executives positioned the products as a response to attackers’ growing use of AI. Mustafa Suleyman, CEO of Microsoft AI, said the new model performed strongly on a cybersecurity benchmark compared with rival systems, while Hayete Gallot, Microsoft’s vice president for security, framed Perception as a way for enterprises to match the speed and scale of AI-enabled threats. Dave Weston, the lead engineer for Perception, said the platform could compress work that previously required multiple specialists into minutes.

TechCrunch noted that the tools are expected to be available in preview on November 3. The launch places Microsoft in a competitive AI security market that includes Anthropic’s Mythos platform and OpenAI’s Daybreak program, both introduced earlier this year.