From Logs to Decisions: Autonomous AI Agents for Real-Time Kubernetes Threat R... Willem Berroubache

This talk covers the development of an open-source, autonomous AI agent architecture for real-time threat detection and response in cloud-native environments using Kubernetes and Kubeflow. The speaker, Willem Berroubache from Orange, discusses how these agents leverage lightweight protocols like A2A and MCP to collect signals through OpenTelemetry, Falco, and Prometheus, allowing for machine learning models to assess threats such as account takeovers and lateral movements. The integration with policy engines like OPA and Kyverno enables agents to initiate secure remediation actions, while contextual large language models provide explanations for each decision made, ensuring auditability and trust. Attendees will gain practical insights, including deployable AI pipelines and templates for building explainable and scalable security defenses in Kubernetes environments using CNCF and open-source technologies.