Securing agent workloads is a continuous process of monitoring, detection, and verification. In this video, we explore how to manage the security posture of your AI agents using AI protection capabilities in Security Command Center (SCC).
Watch along as Aron demonstrates how to maintain a centralized inventory of your AI assets, including agents and MCP servers, and how to utilize Posture Management to detect misconfigurations. You will see how runtime findings from Model Armor, such as jailbreak attempts or indirect prompt injections, are surfaced directly in the SCC dashboard for unified threat management.
We also cover the essential observability tools required for auditing agentic systems. This includes configuring Cloud Logging to capture agent activity. Finally, we discuss defensive strategies, including prioritizing chokepoints and using Sensitive Data Protection (SDP) discovery to identify exposed secrets.
Resources:
Configure agent activity logging → https://goo.gle/46nnjvE
Learn more about AI protection in SCC → https://goo.gle/4r6cF4S
Learn more about defining a security posture in SCC → https://goo.gle/3ZUeJAR
Assess AI security risks in SCC → https://goo.gle/3Zj8zu3
Enforce CMEK encryption for Vertex AI resources → https://goo.gle/4r13faC
Learn more about CMEK → https://goo.gle/4rErD1Z
Secure Credentials for MCP Access with Secret Manager → https://goo.gle/3ZmkHdM
Learn more about toxic combinations and chokepoints → https://goo.gle/4ttLhPZ
Google Secure AI Framework (SAIF) → https://goo.gle/45VWWgi
Speaker: Aron Eidelman
Products Mentioned: Security Command Center, Model Armor, Cloud Logging, Sensitive Data Protection, Customer-managed encryption keys, Vertex AI, Google’s Secure AI Framework
