Microsoft has officially released its Perception security platform, transitioning it from its preview phase to general availability. Announced during the Ignite 2026 conference, this innovative system is designed to coordinate teams of AI agents for various security operations, enhancing Microsoft Defender’s capabilities.
Key Features of Perception Platform
Initially introduced as Project Perception in July, the platform has been tested in public preview since August. The system utilizes playbooks to guide three distinct AI agent teams—red, blue, and green—each tasked with specific security roles. Red agents simulate attackers to identify vulnerabilities, while blue agents are responsible for detection, alert triage, and investigation. Green agents focus on remediation of identified issues.
Microsoft has made four blue team agents widely available, which specialize in alert triage, attack investigation, detection authoring, and threat intelligence. According to Microsoft, the Alert Triage Agent is capable of managing approximately 80% of alerts within a Defender queue, based on data from around 8 million alerts across its security products.
Expanded Agent Capabilities
Furthermore, the Recon Agent from the red team is now generally accessible, and a Web App Pen Test Agent is currently in preview. The platform’s green team has introduced two new playbooks that operate across Microsoft, AWS, and Google Cloud. These include the Posture Prioritization agent, available for general use, and the Risk Reduction agent, which is in public preview.
The Risk Reduction Agent offers temporary vulnerability shielding for assets awaiting a permanent fix, providing an interim layer of protection. This feature ensures that organizations can manage vulnerabilities effectively, even in the absence of immediate patches.
Integrations and Future Outlook
Microsoft is also exploring integrations with various security and management tools, connecting Perception to SIEM, SOAR, and IT service management solutions via MCP and APIs. These integrations are expected to streamline security operations by automating responses and enhancing coordination.
Microsoft has addressed concerns regarding the potential risks of AI agents by monitoring and managing access through Agent 365 and Entra Suite. These tools can identify and manage 60 types of local AI agents, such as Claude Code, OpenAI Codex, and GitHub Copilot, ensuring they remain compliant with organizational policies.
As Microsoft continues to innovate in the cybersecurity domain, the Perception platform represents a significant advancement in AI-driven security measures, promising to enhance defense mechanisms across various cloud environments.
