Microsoft claims a new AI model and Perception platform will significantly cut vulnerability triage from hours to minutes. Early preview starts November 3.
At the start of the week, Microsoft unveiled its first cybersecurity-specialized model product MAI-Cyber-1-Flash and a new AI platform for cyber defense during a small event in San Francisco, strengthening its position in the race for leadership with Anthropic, Google, and OpenAI.
According to descriptions, MAI-Cyber-1-Flash is a model designed ‘to detect complex vulnerabilities in large codebases.’ It is designed to activate MDASH – Microsoft’s hub intended to identify and fix software vulnerabilities.
The new security platform is named Perception and is intended to deploy agent teams to support and automate various workflows in cyber defense, including detection and remediation of bugs. The platform can also integrate with MDASH.
The company claims that MAI-Cyber-1-Flash is significantly more powerful and cost-efficient than competing models based on metrics in the well-known AI cybersecurity benchmark.
Key Takeaways and Market Context in Cybersecurity
We are extremely pleased to report our results.
– Microsoft
We have MAI-Cyber-1-Flash, connected with GPT-5.4 inside the MDASH hub – ahead of Gemini, GPT-5.5 Cyber, GPT-5.6 Sol, and Mythos 5 on Cyber Gym, the main benchmark we all use. A gold standard?
– Microsoft
We are rolling this out into production right away,
– Microsoft
Perception is a tool for enterprise defenders to defend against AI with the scale and speed inherent to attacks.
– Microsoft
We have significantly increased effectiveness thanks to red, blue, and green agent teams: red teams simulate possible attacks, blue teams detect and classify bugs, and green teams implement fixes.
– Microsoft
We have moved from hours of manual work by a few specialists to minutes, and now we have a solution for everything: from detection and prioritization to remediation and even code fixes.
– Microsoft
While AI opens new defensive capabilities for companies, its availability also raises risks: malicious actors could deploy similar tools for threats, jeopardizing data and system protection.
According to Microsoft’s plans, the new security tools will be available in a preview on November 3 amid growing competition in the AI cybersecurity field. Earlier this year, Anthropic launched Mythos – a security platform that was available to a small cohort of partners through the Glasswing program, and OpenAI introduced Day Break in May.
In the context of the future development of this niche, Microsoft plans to strengthen its position by integrating Perception with MDASH and using MAI-Cyber-1-Flash in production, while also introducing new approaches to threat management and incident-response automation.
