Microsoft aims to make AI security cheaper and more efficient with a new platform designed to compete with Anthropic Mythos.
According to The Information, Microsoft is working on a new AI security solution under the codename Project Perception. The service is intended to help organizations detect and remediate vulnerabilities in their IT environments, and would compete with Anthropic Mythos.
According to The Information, Microsoft primarily wants to differentiate itself with a lower price point. While Anthropic Mythos uses powerful but expensive AI models, Project Perception would combine various models and automatically deploy the most suitable model depending on the task.
Multiple AI models for a single security platform
Project Perception would combine AI technology from Microsoft, OpenAI, and Anthropic to detect, analyze, and, where possible, suggest solutions for vulnerabilities.
Instead of having all tasks processed by the same model, the platform automatically selects the most appropriate model per task. This allows less complex analyses to be performed by cheaper models, while heavier security issues are sent to more powerful models. In this way, Microsoft aims to reduce operational costs without sacrificing performance.
Microsoft focuses heavily on the enterprise market
Microsoft plans to reintroduce its own advanced AI models next year and is also investing in specialized AI hardware. At the same time, the company is positioning itself more emphatically in the areas of security and governance—domains where Microsoft already holds a strong position with business customers thanks to Windows, Microsoft 365, and Azure.
According to CEO Satya Nadella, companies increasingly expect guarantees regarding how their data is processed. With Project Perception, Microsoft appears to be extending that broader AI strategy to security, a market in which it seeks to differentiate itself more clearly from competitors.
From user to competitor
Project Perception appears primarily intended as an alternative to Anthropic Mythos, an AI platform that automatically detects vulnerabilities in corporate environments and is being monitored by the US government. The most powerful Mythos models were temporarily unavailable outside the United States in recent months due to US export restrictions. Although those restrictions have since been lifted, they highlighted how dependent organizations can become on a single AI provider.
Goodbye Mythos, hello Project Perception
The fact that Microsoft is now working on Project Perception is not entirely unexpected. The company previously tested Claude Mythos to detect vulnerabilities in software faster during development. At the time, Microsoft considered Anthropic’s security model a benchmark for AI-driven vulnerability detection. Since then, the strategy seems to have shifted. Instead of relying exclusively on one AI provider, Microsoft is now developing its own platform that combines multiple models. Depending on the task, Project Perception automatically selects a model from Microsoft, OpenAI, or Anthropic.
This approach is intended not only to reduce operational costs but also to limit dependence on a single AI vendor. As such, Project Perception appears to be more than just a new security service: it is a combination of AI, security, and governance for enterprise customers.