Enhancing Agent Security with Real-Time Protection in Microsoft Copilot Studio

As AI agents become integral to critical business workflows, the demand for robust security measures increases. Microsoft Copilot Studio already offers strong built-in protections against agent manipulation, but for organizations requiring deeper oversight and proactive control, a new feature is now available in public preview. This feature provides advanced near-real-time protection during agent runtime for enhanced security.

This capability enhances AI agent security by allowing organizations to integrate their own monitoring systems, such as Microsoft Defender, as well as security platforms from other providers or custom-built tools. These integrations enable near-real-time evaluation and control of agent behavior during runtime.

When connected, external systems become part of the agent's decision-making process. They can block unsafe actions, even if the agent intends to execute them. For instance, if an external system detects that the agent plans to send an email oversharing information, it can block the email from being sent. Administrators can apply these protections across multiple agents and environments using the Power Platform Admin Center.

AI agents face unique threats, such as cross prompt injection attacks (XPIA), where malicious prompts trick agents into leaking data or misusing tools. Copilot Studio includes default protections against both XPIA and user prompt injection attacks (UPIA). These defenses block suspicious prompts in near-real time, reducing the risk of data loss or unauthorized actions.

For organizations with advanced security needs, built-in protections may not suffice. That's where near-real-time protection provides an additional layer of defense. With advanced runtime protection, Copilot Studio calls the connected security system during the agent's runtime. The system reviews the agent's planned actions and decides whether to approve or block them.

This setup offers stronger control over agent behavior while preserving a smooth user experience. It supports a "bring your own protection" model, allowing integration with various security systems. This flexibility helps organizations align AI agent security with internal policies, industry standards, and regional compliance.

In addition to blocking threats before they occur, Copilot Studio creates detailed audit logs for every interaction with the external system. Administrators can use these logs to track attempted breaches, identify vulnerable agents, and improve future deployments. These logs also help evaluate the performance of the external monitoring system.

Advanced security for AI agents is essential as agents increasingly handle complex and sensitive tasks. External monitoring provides tools to enforce compliance, detect and block threats, and gain visibility without compromising performance. This new capability in Copilot Studio empowers organizations to control their AI agent security strategy effectively.