Session
Who Authorized That Agent? Accountability and Identity in Agent-Native Kubernetes
Your team deploys an agent for operations. It needs kubectl, cloud access, observability tools. You create a service account with broad permissions.
Now the agent has more access than the engineers.
Service accounts were designed for predictable services. Agents decide actions dynamically - a fundamental mismatch.
Three security gaps:
Identity Attribution: Audit logs show the service account, not which user triggered the agent or why.
Permission Scope: Agents need broad capability but should execute with task-specific permissions. Static RBAC can't express this.
Tool Chain Opacity: Agents invoke MCP servers dynamically. Each needs different permissions; Kubernetes sees one identity.
This talk shows an MCP server on Kubernetes: user context propagation, per-tool RBAC via admission controllers, and traces linking requests to decisions.
Your agents have the keys. This talk shows you the patterns to track who authorized them, what they decided, and why - before it's too late.
Atulpriya Sharma
Principal Developer Advocate @ Improving | CNCF Ambassador
Hyderābād, India
Links
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top