Microsoft is preparing to launch a new AI-powered cybersecurity product that could change how companies find and fix software vulnerabilities before attackers exploit them.
The product, known internally as Project Perception, is designed to scan enterprise systems for security flaws, recommend fixes, and in some cases automatically apply them. What makes it different is that it does not rely on a single AI model. Instead, it combines models from Microsoft, OpenAI, and Anthropic, choosing the best one for each task.
The reported product arrives as businesses face growing pressure to defend against AI-assisted cyberattacks. Security teams are increasingly looking for tools that can identify weaknesses faster than human analysts can, especially as advanced AI models become more capable of discovering software bugs and developing exploits.
A Multi-Model Approach to Enterprise Security
According to reports, Project Perception uses what Microsoft calls a model router. Rather than sending every request to one AI model, the router selects between Microsoft’s own models and models from OpenAI and Anthropic depending on the type of security task being performed.
This approach is expected to improve both performance and cost by matching each job with the most suitable model.
The system is built to inspect enterprise code, applications, and infrastructure for vulnerabilities.
It can identify potential security issues, suggest patches, and automatically remediate some of those weaknesses.
Microsoft reportedly hopes to offer the product at a lower price than Anthropic’s Mythos, which has become known for its advanced vulnerability discovery capabilities but is also considered expensive to operate.
Why Microsoft is Entering this Race
Project Perception reflects a broader shift in cybersecurity. AI is no longer being used only to defend systems. It is also giving attackers new ways to search for weaknesses at scale. This has created demand for defensive AI tools that can move just as quickly.
Anthropic recently demonstrated how powerful AI can become in cybersecurity with its unreleased Mythos model, which the company says discovered thousands of previously unknown vulnerabilities across major operating systems and web browsers. Access to that model remains tightly restricted because of its advanced capabilities.
Microsoft now appears to be building its own answer to that challenge. Instead of creating a single security model, it is combining multiple frontier models into one platform that can adapt to different workloads.
This strategy also aligns with Microsoft’s wider push toward giving enterprise customers access to multiple AI providers rather than relying only on OpenAI.
What It Means for Businesses
If Microsoft launches Project Perception as reported, companies could gain a tool that continuously searches for security weaknesses before hackers have the opportunity to exploit them. It could also reduce the workload on security teams by automating parts of vulnerability detection and remediation.
The product has not yet been officially announced by Microsoft, but reports indicate it could debut soon as the company strengthens its AI security portfolio. As AI continues to reshape both cyber defense and cybercrime, tools that proactively identify vulnerabilities are becoming an increasingly important part of enterprise security.
