Microsoft launches MAI-Cyber-1-Flash and a new security platform

In short: Microsoft has launched its first cybersecurity-focused AI model and a new security platform called Perception, with a public preview planned for November 3.

What happened

Microsoft announced a new AI model called MAI-Cyber-1-Flash at an event in San Francisco. The company says the model is built to find hard-to-spot security holes in large and complicated software projects.

Microsoft said the model powers MDASH, its internal system for finding and fixing software vulnerabilities (weak spots in code that hackers can use, like an unlocked window). Microsoft also said MAI-Cyber-1-Flash performed well on a common cybersecurity test used to compare AI models.

Alongside the model, Microsoft introduced a new security platform called Perception. Perception is designed to run “teams” of AI agents, which are AI helpers that can take steps on their own, like following a checklist without being asked every time. Microsoft says these agents can simulate attacks, spot bugs, sort them by urgency, and help apply fixes, and it can integrate with MDASH.

Microsoft described three agent roles. “Red teams” simulate how an attacker might break in, “blue teams” look for and organize existing problems, and “green teams” take corrective actions to address them.

Why it matters

Cyberattacks are growing, and Microsoft says attackers are also using AI to move faster. Tools like Perception are aimed at helping companies respond more quickly, since security work often takes many people and many hours. This announcement also adds pressure in a crowded market, where companies like Anthropic and OpenAI have launched their own AI security products.

Source: TechCrunch AI