The AI control gap: Securing inference and agents on Kubernetes

As AI workloads like inference and autonomous agents move to Kubernetes, they introduce a massive control gap. This talk explores how to use Kubernetes admission control as a unified policy layer to enforce guardrails and secure the AI supply chain. Nirmata co-founder Ritesh Patel examines the AI control gap emerging as enterprises deploy generative AI, RAG pipelines, and autonomous agents on Kubernetes. He demonstrates why post-deployment detection is not enough and how to use Kubernetes-native policy, Kyverno, to enforce security by design.

• Understand vulnerabilities across the AI supply chain, inference, and agent planes.
• Learn how to use admission control to block over-privileged AI agents and unverified models.
• Explore strategies for building provable, automated compliance for AI workloads.