Microsoft is developing Project Perception, an AI-powered cybersecurity tool intended to rival Anthropic’s Mythos in vulnerability detection and remediation. The tool routes tasks across multiple language models from Microsoft, OpenAI, and Anthropic, assigning routine work to cheaper models while reserving premium models for complex reasoning, to make continuous vulnerability management more affordable than Mythos. It is expected to launch this month and can scan a company’s code, cloud infrastructure, and endpoints to identify exploitable weaknesses and propose fixes, though it raises data governance questions about which external models process sensitive information.
