Microsoft Launches AI Cybersecurity Model and Agentic Platform

At a recent event in San Francisco, Microsoft introduced its first AI model for cybersecurity called MAI-Cyber-1-Flash, along with a new security platform named Perception. This model is designed to spot weaknesses in complex code and works in tandem with MDASH, a Microsoft tool for detecting and fixing vulnerabilities. Perception employs AI teams—red, blue, and green—to simulate attacks, find problems, and apply solutions. Microsoft claims that MAI-Cyber-1-Flash outperforms competitors like Gemini and GPT 5.5 Cyber in the Cyber Gym benchmark. Mustafa Suleyman, CEO of Microsoft AI, announced its immediate availability, while Hayete Gallot emphasized AI's role in combating AI threats. The tools will be previewed starting November 3.