Microsoft has introduced “Project Perception,” an agentic security system intended to help defend organizations against cybersecurity threats enabled by artificial intelligence. The company says the system combines multiple types of inputs—signals and context—with AI models and specialized agents to detect, prioritize, and respond to suspicious activity. Microsoft describes the approach as turning threat signals into real-time protections, using AI to help counter threats that may themselves be generated or accelerated by AI.
According to Microsoft’s announcement, Project Perception coordinates three classes of specialized agents. The “Red team” agents focus on identifying potential exploitation paths that could allow a bad actor to compromise an organization’s technology architecture. Microsoft also describes the overall framework as reasoning over relevant information and taking autonomous action as needed, rather than relying only on static rules.
Both sources frame the rollout as part of Microsoft’s broader effort to address the security implications of AI-powered attacks by using an “agentic” model that can understand context and execute defensive steps.
