Microsoft Launches First Cybersecurity AI Model and Agentic Security Platform

Microsoft unveiled its first cybersecurity-specialized AI model and a new agentic security platform at a small event in San Francisco on Monday, July 28, 2026, entering a field that already includes competing offerings from Anthropic, Google, and OpenAI.

The model, called MAI-Cyber-1-Flash, is built to identify vulnerabilities in complex codebases and powers MDASH, Microsoft’s software vulnerability identification and remediation harness. The accompanying platform, Perception, deploys teams of AI agents to assist with and automate security workflows, including bug detection and remediation, and can also integrate with MDASH.

Mustafa Suleyman, CEO of Microsoft AI and co-founder of DeepMind, said the model outperforms rivals on Cyber Gym, a widely used AI cybersecurity benchmark. “We have MAI-1 Cyber Flash binded with GPT 5.4 inside of the MDASH harness — which beats out Gemini, GPT 5.5 Cyber, GPT 5.6 Sol, and Mythos 5 on Cyber Gym, which is the primary benchmark that we all use,” Suleyman said, adding that the model is being shipped to production immediately.

Perception organizes agents into three teams: red teams that simulate potential attacks and identify likely vulnerabilities, blue teams that detect and triage existing bugs, and green teams that take corrective action against them. Dave Weston, the platform’s lead engineer, said the system compresses what previously required hours of manual work from multiple security specialists into minutes, delivering detection, posture fixes, and code fixes in a single workflow.

Hayete Gallot, Microsoft’s vice president for security, framed the platform as a direct response to the growing use of AI by cybercriminals, describing Perception as a way for enterprise defenders to “defend against AI with AI at the scale and speed that the attackers have.”

Both MAI-Cyber-1-Flash and Perception are set to enter preview on November 3. They will compete with Anthropic’s Mythos security platform, released earlier through a program called Glasswing, and OpenAI’s security solution launched in May through a program called Daybreak.

Source: TechCrunch

This article was generated by AI and cites original sources.
Scroll to Top