Microsoft unveiled its first cybersecurity-specialized AI model alongside a new AI-powered security platform at a small event in San Francisco on Monday. The announcement signals a direct challenge to major competitors in the space, including Anthropic, Google, and OpenAI.
MAI-Cyber-1-Flash: A Model for Vulnerability Discovery
The new model, named MAI-Cyber-1-Flash, is designed to “find challenging vulnerabilities in complex codebases,” according to Microsoft. It integrates with MDASH, Microsoft's harness for software vulnerability identification and remediation. The company claims that MAI-Cyber-1-Flash is significantly more powerful and cost-effective than rival models, based on its performance on the established Cyber Gym benchmark.
Mustafa Suleyman, co-founder of DeepMind and current CEO of Microsoft AI, expressed excitement about the results. “We have MAI-Cyber-1-Flash bound 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, the golden benchmark,” he said. “We’re shipping this into production immediately.”
Perception: An Agentic Security Platform
Microsoft also launched Perception, an AI platform that deploys teams of agents to automate and assist with security workflows, including identifying and remediating bugs. Perception integrates with MDASH and is built to help enterprise defenders “defend against AI with AI at the scale and speed that the attackers have,” said Hayete Gallot, Microsoft’s vice president for security.
Perception employs three types of agent teams:
- Red teams simulate potential attacks, providing detailed context about threat actors and likely vulnerabilities.
- Blue teams focus on detecting and triaging existing bugs.
- Green teams take corrective actions to fix those bugs.
Dave Weston, lead engineer for Perception, highlighted the efficiency gains. “We’ve gone from this taking hours and hours of manual work from multiple specialized folks across the security organization — appsec hunters, remediation engineers, you name it — and in minutes, we have a fix for all of this. Not only do we discover the issues and prioritize them, but we have detection, posture fixing, and even a code fix.”
The Growing AI Cybersecurity Landscape
Microsoft’s new tools, which will be available in preview starting November 3, 2026, enter an increasingly crowded field. Earlier this year, Anthropic launched Mythos, a security platform released to a select group of partners through its Glasswing program. OpenAI has also introduced its own security-focused AI tools, intensifying competition in the rapidly evolving market.
As hackers increasingly leverage AI in their attacks, Microsoft’s latest offerings aim to provide defenders with advanced, automated capabilities to counter emerging threats. The company’s move underscores the growing importance of AI-driven cybersecurity in an era where both attackers and defenders are racing to harness the technology.
via TechCrunch AI