Microsoft launches MAI-Cyber-1-Flash and Perception security platform
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Microsoft launched MAI-Cyber-1-Flash, its first cybersecurity-specialized model designed to find vulnerabilities in complex codebases, alongside Perception, an agentic security platform that deploys red, blue, and green teams to automate threat detection and remediation.
- Mustafa Suleyman (CEO of Microsoft AI) claimed MAI-Cyber-1-Flash combined with GPT 5.4 in Microsoft's MDASH harness beats Gemini, GPT 5.5 Cyber, GPT 5.6 Sol, and Mythos 5 on the Cyber Gym benchmark, which he called 'the golden benchmark.'
- Perception is shipping into production immediately and will be available in preview on November 3; lead engineer Dave Weston said the platform reduces hours of manual work from multiple security specialists to minutes of automated detection, prioritization, and code fixes.
- Hayete Gallot (Microsoft VP for security) framed Perception as a way for enterprise defenders to 'defend against AI with AI at the scale and speed that the attackers have,' as AI-powered cyberattacks accelerate.
- The launch explicitly targets Anthropic (which released Mythos through a limited Glasswing program), Google, and OpenAI (which launched Daybreak in May), entering a crowded field of AI cybersecurity solutions.
Why it matters: Microsoft is stepping into a space where Anthropic's Mythos and OpenAI's Daybreak remain in limited release, and Suleyman's claim of Cyber Gym supremacy—if validated—would give enterprise security teams an integrated alternative that compresses hours of manual work into minutes of automated detection and code fixes.
