← /editions
#125

Agents Get the Keys: Identity Becomes the Control Plane

// Anugal brings agentic AI into identity governance and access decisions while agents broadly move from assisting to executing in the enterprise, making identity the control plane for agentic systems.

$ today_in_10_seconds

Agentic AI is moving into the two places enterprises guard most closely: who gets access to what, and what actually gets done. Anugal is putting agents into identity governance decisions, and SiliconANGLE describes the wider shift from pilots to execution. Agents can already act. The open question is who governs them when they do.

$ lead_story

Anugal Brings Agentic AI Into Enterprise Identity Governance and Access Decisions

Anugal is applying agentic AI to identity governance and access decisions. These are the approve, deny, and revoke workflows that usually crawl through ticket queues and quarterly access reviews. They suit agents well because they're multi-step, context-heavy, and rule-bound, and humans mostly rubber-stamp them today. The catch is that an agent making access decisions is itself a privileged identity, which raises the stakes on audit trails, explainability, and reversibility. If you're evaluating this or building something like it, find out what the agent decides on its own versus only recommends, and how each decision is logged.

// why_it_matters: When agents start deciding who gets access, agent governance stops being theoretical.

read more →

$ also_worth_attention

  • Agentic AI moves into enterprise execution

    SiliconANGLE describes the shift from agents that assist humans to agents that complete work inside enterprise systems. Execution means write access to production systems, so the bottleneck moves away from model capability and toward permissions, observability, and rollback.

    link →

$ bigger_picture

The trend behind the headlines

Today's two stories are really one story. Every agent that executes needs credentials, every credential needs governance, and governance is now being handed to agents too. That makes identity the control plane for agentic systems. The model is no longer the hard part. The standing service account with broad permissions, the default for most agent deployments, won't survive contact with auditors. Expect the winning pattern to be scoped, short-lived, per-task permissions with a human-readable reason attached to every action. For founders, the opportunity is less in smarter agents and more in the plumbing that lets enterprises trust them.

$ tool_of_the_day

OpenFGA

OpenFGA is an open-source, fine-grained authorization engine inspired by Google's Zanzibar and hosted by the CNCF. You can use it to model what an agent may do and on whose behalf, and to check each action at runtime instead of baking permissions into a service account. If your agents are about to get write access to anything, this is a solid foundation.

explore OpenFGA →

$ subscribe --free

Get the next edition in your inbox

// DAIly Agent covers agentic AI three times a week

// free, no spam, exit anytime