Microsoft just made its move on the AI security market. On Monday, the company launched its first cybersecurity-specialized model, MAI-Cyber-1-Flash, alongside a new agentic security platform called Perception, according to TechCrunch AI. The launch happened at a small event in San Francisco, and Microsoft didn’t hide who it’s aiming at. This is a direct swipe at Anthropic, Google, and OpenAI.
What stands out here is the ambition. Microsoft AI CEO Mustafa Suleyman, the DeepMind co-founder, said the company is “shipping this into production immediately.” That’s an aggressive posture in a field that usually launches quietly to a handful of partners.
Here’s what Microsoft actually announced.
- MAI-Cyber-1-Flash, a model built to hunt bugs. Microsoft describes it as a model built “to find challenging vulnerabilities in complex codebases.” It powers MDASH, the company’s harness for spotting and fixing software vulnerabilities. Per TechCrunch AI, Microsoft claims it’s both more powerful and more cost-effective than rival models.
- Perception, an agentic security platform. This is the bigger story. Perception deploys teams of AI agents to run and automate security workflows, from finding bugs to fixing them. It can also plug into MDASH. Microsoft’s VP for security, Hayete Gallot, framed it as a way for enterprise defenders to “defend against AI with AI at the scale and speed that the attackers have.”
- Red, blue, and green agent teams. Perception splits the work across three agent squads. Red teams simulate attacks, including context on likely threat actors and the vulnerabilities they’d target. Blue teams detect and triage existing bugs. Green teams take “corrective actions” to fix them. It’s the classic security team structure, rebuilt with agents.
- A benchmark claim aimed squarely at competitors. Suleyman said the MAI-1 Cyber Flash model, paired with GPT 5.4 inside the MDASH harness, “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. The golden benchmark.” In plain terms, Microsoft is claiming top scores on the test the industry treats as the standard.
Why the speed matters
The pitch isn’t just detection. It’s collapse of time. Dave Weston, the lead engineer on Perception, put it bluntly. “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,” he said. He added that the platform doesn’t just find and prioritize issues, it delivers “detection, posture fixing, and even a code fix.”
That’s the real selling point for enterprises. Security teams are expensive and stretched thin, and Microsoft is arguing it can compress a multi-person workflow into a few minutes of agent labor.
Availability and the competition
The tools land in preview on November 3, according to TechCrunch AI. And they’re walking into a crowded room. Anthropic launched a security platform called Mythos earlier this year, released to a small group of partners through a program named Glasswing. OpenAI shipped its own security solution in May under a program called Daybreak. Google is in the mix too.
So Microsoft isn’t first. What it’s betting on is going broad and going fast while rivals stay in limited-access mode.
My take
The timing is the point. As TechCrunch AI notes, attackers are increasingly using AI in their campaigns, which raises the stakes for defenders who can’t keep up manually. Microsoft is reading that moment and positioning Perception as the answer for enterprises that need speed to match the threat.
Two things are worth watching. First, the benchmark claims are Microsoft’s own, tied to Cyber Gym, and independent testing will tell the real story once the preview is in hands. Second, handing agents the power to take “corrective actions” and push code fixes automatically is powerful and a little nerve-racking. Enterprises will want tight guardrails before they let green teams rewrite production code on their own.
November 3 is the date to circle. That’s when we’ll see whether Perception holds up outside the demo room. More details are available at the original TechCrunch AI report.