In this Microsoft Ignite 2025 session, Ramya Chitrakar, Moran Gutman, and Asaf Nakash demonstrate how to secure AI agents using Microsoft Defender, offering live demos and essential security strategies.

Secure Your AI Agents with Microsoft Defender: Best Practices from Ignite 2025

Session Overview

This intermediate session from Microsoft Ignite 2025, presented by Ramya Chitrakar, Moran Gutman, and Asaf Nakash, focuses on enhancing the security and governance of AI agents using Microsoft Security solutions and tools.

Key Topics Covered

  • Rising Attack Surfaces and Threat Vectors: The presenters discuss how increasing reliance on AI agents exposes new vulnerabilities, necessitating advanced security practices.
  • Microsoft Agent 365 as Control Plane: Introduction of Microsoft Agent 365 provides a centralized platform for managing and securing AI agents across cloud environments.
  • Graph Database Insights: Deep dive into mapping attack paths using cloud resource graphs and identifying potential risk areas.
  • New Defender Capabilities: Announcement and demo of three newly launched security features in Microsoft Defender for AI agents.
  • Live Demos:
    • Demonstration of a Java Customer Helper Agent’s operation.
    • Security risk scenarios and attacker simulation setups.
    • Integrating Copilot Studio for responsible AI and automated threat blocking workflows.
    • Defender’s ability to correlate attack attempts and generate security alerts.

Best Practices Outlined

  • Leverage central control planes like Microsoft Agent 365 for unified AI management
  • Use graph-based attack path analysis to proactively identify exposure
  • Employ multi-layered defense strategies, including automation and responsible AI guidance
  • Integrate Microsoft Defender’s latest features for real-time detection and response
  • Combine Copilot Studio with Defender for robust AI system governance

Additional Resources

Conclusion

By implementing the practices and capabilities demonstrated, organizations can bolster the resilience of their AI agents and confidently manage emerging security risks in the cloud.