Your AI Is a Toddler With Root Access: Why Kubernetes Policies Are Your Last Line...
summary
Claim: AI agents can rapidly create and deploy misconfigured resources so Kubernetes admission policies must be enforced as last-line blast doors. Demo builds a searchable vector database of company rules and CRD summaries, uses Cloud Code and an MCP agent to semantically map intents to ~20–30 generated Kyverno policies from ~1,000 cluster schemas (export step ~1 hour). Generated policies are applied to the cluster and enforce region and resource constraints successfully; violations are rejected at admission.
AI-written summary · about
chapters (7)
- 0:00Kubernetes policies as rules
- 2:26Policies as last defense
- 5:56Demo setup and policy creation
- 9:27Generating OPA policies
- 13:21Ideal policy enforcement
- 17:45Semantic search enforcement
- 20:20Question and answer session
related talks
25mWhat We Learned Securing AI Agents at Scale on Multi-Tenant Kubernetes Clusters
35mHow to fight misconfiguration and bad intentions with Kubernetes admission policy
30mRedefining Access Control: Scaling Policy as Code for Humans and AI Agents
35mShift-Left Policy Validation in Kubernetes With Kyvern...
44mMCP Security: Keep Your AI Agents from Spilling the Tea
32mCloud Native Policy as Code: Principles and Best Practices
36mCentralized Policy Management
20mTrust but Verify