Identity in the Agentic Era

Reading Time: < 1 minute

Identity in the Agentic Era

AI agents are non-human identities, and most teams are repeating every IAM anti-pattern from the cloud-native era — faster, with actors that are autonomous and manipulable. A 6-episode series applying a decade of IAM and Zero Trust practice to agentic AI architecture.

Who it’s for: Security and platform engineers who already understand cloud IAM (roles, workload identity, least privilege) and want to see exactly where those patterns break — or need reinforcing — once the caller is an LLM agent instead of a human or a service account.

Format: Each episode connects a specific IAM/access-control failure mode to its agentic-AI equivalent — RAG access control, OIDC workload identity for LLM pipelines, IAM abuse via prompt injection, audit trails for agent actions, and zero trust mutual authentication for agent-to-agent calls.


Published Episodes

EP Title
EP01 The Non-Human Identity Problem Is Back
EP02 RAG Access Control: The IAM Layer Your Vector Database Doesn’t Have
EP03 OIDC and Workload Identity for LLM Pipelines

Coming Up

EP Title
EP04 When Prompt Injection Becomes IAM Abuse
EP05 Auditing AI Agent Actions: What CloudTrail Tells You and What It Misses
EP06 Zero Trust for AI Agents: Mutual Authentication in Agentic Architectures

Start with EP01: The Non-Human Identity Problem Is Back →

Get new episodes in your inbox → subscribe