Webinar Description
Key Takeaways
- Focuses on security and governance challenges specific to autonomous AI agents operating in production environments
- Addresses identity management, authorization controls and auditability requirements for agentic systems
- Designed for solutions architects, security architects, platform engineers and CISOs working in regulated industries
- Covers architectural constraints and security patterns for Kubernetes-based infrastructure
- Presented by Tigera, with relevance to organisations running workloads on AWS EKS, Azure AKS, Google GKE and other managed Kubernetes platforms
Introduction
Securing Autonomous AI Agents in Production is a technical webinar examining the security and governance requirements that emerge when organisations deploy autonomous AI agents across enterprise infrastructure. Led by Ivan Sharamok, Principal Solutions Architect at Tigera, the session targets technical decision-makers responsible for securing AI workloads in regulated environments. As enterprises progress beyond initial concerns about large language model accuracy, attention is shifting toward the operational risks introduced when AI agents act autonomously within critical systems.
The Security Challenge of Autonomous AI Agents
Autonomous AI agents represent a significant departure from traditional software components. Unlike conventional applications that execute predefined logic, agentic systems make decisions and take actions with varying degrees of independence. This autonomy introduces security considerations that existing frameworks were not designed to address. When an AI agent can initiate network connections, access data stores or invoke APIs without direct human oversight, the attack surface expands considerably.
The webinar addresses three interconnected risk areas that become particularly acute in production deployments. Identity management for autonomous agents requires rethinking how systems authenticate and track non-human actors that may spawn multiple processes or assume different roles during execution. Authorization controls must balance operational flexibility against the principle of least privilege, ensuring agents can perform necessary tasks without accumulating excessive permissions. Auditability becomes essential for compliance and incident response, requiring comprehensive logging of agent decisions and actions in formats that support forensic analysis.
Architectural Patterns for Regulated Environments
Organisations operating in regulated sectors face additional complexity when deploying autonomous AI agents. Financial services, healthcare and other industries subject to strict compliance requirements must demonstrate that AI systems operate within defined boundaries and that their actions can be traced and explained. The session explores architectural constraints that enable organisations to run agentic systems while maintaining the governance posture required by regulators.
A central concept discussed is the implementation of a dedicated control plane for autonomous agents. This architectural approach provides centralised visibility and policy enforcement across distributed agent deployments, enabling security teams to monitor behaviour patterns, detect anomalies and enforce consistent controls regardless of where agents operate within the infrastructure.
Kubernetes as the Foundation for Agent Security
The webinar situates autonomous AI agent security within the broader context of Kubernetes and cloud-native infrastructure. Many organisations deploying AI workloads at scale rely on managed Kubernetes services including AWS EKS, Azure AKS, Google GKE, Red Hat OpenShift, SUSE Rancher and Mirantis. These platforms provide the orchestration capabilities necessary for running distributed AI systems but require additional security layers to address the unique characteristics of autonomous agents.
Tigera’s Calico platform provides network security and observability for Kubernetes environments, while Lynx extends these capabilities specifically for AI agent workloads. The relationship between container orchestration, network policy enforcement and agent-specific security controls illustrates how organisations must layer multiple technologies to achieve comprehensive protection.
Who Should Attend
The session is designed for technical leaders responsible for infrastructure security and AI deployment decisions. Solutions architects evaluating how to integrate autonomous agents into existing systems will find relevant guidance on architectural patterns. Security architects and CISOs concerned with governance and compliance will benefit from the focus on auditability and control mechanisms. DevOps and platform engineers tasked with implementing and operating AI workloads on Kubernetes will gain practical insights into security configurations and monitoring approaches.
Organisations in regulated industries considering or actively deploying autonomous AI agents represent the primary audience, though the security principles discussed apply broadly to any enterprise running agentic systems in production.