The three layers of agentic AI security: A defense-in-depth architecture for autonomous agents
Presented by Nutanix Autonomous systems that can reason, make their own decisions, and execute actions across an environment introduce a category of risk that application-level controls were never built to contain. Treating that risk as a single problem produces incomplete architectures, says Oscar Wahlberg, senior director of product management at Nutanix. "The guardrails to catch a malicious…
Nutanix's Oscar Wahlberg highlights the need for a defense-in-depth architecture when deploying autonomous agents that can reason and act autonomously. Treating security as a single problem fails to address the unique risks posed by these agents. The solution involves dividing security responsibilities across three layers.
First, the infrastructure layer focuses on establishing trust in the environment where AI agents run. Technologies such as platform attestation, confidential computing, and secure boot verify the authenticity of agents. Isolating AI production workloads within assigned scopes helps prevent risks like model tampering and unauthorized access to sensitive data.
Next, the network layer governs communication among AI agents, APIs, applications, and enterprise systems. As agents interact with various components, they create dynamic, complex communication patterns that traditional static network configurations cannot handle. By treating AI agents as a new network identity and enforcing dynamic policies, organizations can prevent lateral movement, data exfiltration, and unauthorized access.
Finally, the control plane layer serves as the centralized decision-making hub for managing agent permissions, tool access, resource consumption, and runtime visibility. A unified control plane helps enforce consistent policies across all agents, mitigating risks like privilege misuse, runaway agents, and excessive model consumption. Nutanix's Agent Gateway acts as a single endpoint for managing agents, providing governance capabilities and integrating with networking vendors to ensure secure interactions.
In summary, a defense-in-depth architecture for autonomous agents requires a layered approach that addresses distinct categories of risk. By establishing trust at the infrastructure layer, governing network communications, and centralizing control at the control plane layer, organizations can create a robust security framework to protect their AI environments.
Written by urgent.news from VentureBeat's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
