In July 2026, Microsoft unveiled its latest AI-driven security offering, 'Project Perception.' This innovative product harnesses the power of multiple proprietary cybersecurity AI models, positioning itself as a budget-friendly substitute for Anthropic's renowned Mythos model. Embracing a multi-model collaborative framework, Project Perception seamlessly integrates AI models sourced from Microsoft, OpenAI, and Anthropic. It employs sophisticated 'model routing' technology to adeptly distribute tasks, ultimately slashing operational expenses.

Microsoft's in-house model, MAI-Cyber-1-Flash, rivals the performance of Mythos in specific tasks, all while maintaining a notably lower cost. Project Perception is engineered to facilitate code analysis, pinpoint vulnerabilities, and execute automated fixes. Anticipation mounts as it is slated for a public preview launch on August 3.

In a move to foster broader industry collaboration, Microsoft has also made its Agent SDK and policy template library accessible. The inaugural batch of integrations encompasses over 83 independent software vendors spanning diverse sectors, including finance, healthcare, and manufacturing.