Who should own governance when access spans humans, service accounts, and AI agents?

Why This Matters for Security Teams

Ownership becomes difficult when one access model covers people, service accounts, and AI agents, because each behaves differently and is controlled by different teams. Human identities fit HR-backed joiner-mover-leaver processes, service accounts usually sit inside platform or application ownership, and AI agents introduce runtime decisions that can change from task to task. The governance question is less about who clicks the button and more about who defines the policy and accepts the risk. That is why current guidance increasingly points to a single policy model with actor-specific enforcement, rather than three separate standards that drift apart. NIST’s NIST Cybersecurity Framework 2.0 is useful here because it frames identity as an enterprise governance issue, not a siloed control. NHIMG research on The State of Non-Human Identity Security shows how often NHI controls lag human identity controls, which is exactly what happens when ownership is fragmented. In practice, many security teams discover that governance gaps only surface after access sprawl, not through intentional design reviews.

How It Works in Practice

The cleanest operating model is a three-layer split: enterprise policy, control ownership, and technical execution. Identity governance or the identity programme should own the policy standard, because it is the only function positioned to apply one access principle consistently across humans, service accounts, and AI agents. IAM and PAM then implement the human and privileged workflow controls, while NHI governance and security engineering handle secrets, workload identity, rotation, and runtime telemetry for non-human actors. For agents specifically, the control problem is not static membership but context at execution time, so policy must be evaluated when the request occurs, not only when the account is created. That is where intent-aware authorization, JIT credentials, and workload identity become practical, rather than theoretical. Frameworks such as the OWASP Agentic AI Top 10 and NIST AI Risk Management Framework both reinforce that autonomous systems need governance at runtime, not just a provisioning checklist. NHIMG’s lifecycle guidance for managing NHIs is useful for defining where secrets, rotation, and revocation sit in the operating model.

• Use one policy authority to define least privilege, approval thresholds, and separation of duties.
• Assign execution owners by actor type: IAM for humans, PAM for privileged elevation, platform or app owners for service accounts, and security engineering for agent tooling and secrets.
• Require workload identity and short-lived credentials for non-human access wherever possible.
• Log the actor, the context, and the decision so reviews can distinguish entitlement drift from legitimate automation.

These controls tend to break down in highly federated enterprises where each application team provisions its own identities and no single group can enforce revocation or policy consistency.

Common Variations and Edge Cases

Tighter central governance often increases operational overhead, so organisations need to balance speed against control assurance. In highly regulated environments, it is common for IAM, PAM, and NHI ownership to be split by compliance boundary, but best practice is evolving toward a single governing authority with delegated execution. That matters because a service account used by a pipeline and an AI agent with tool access can both become privilege escalation paths if ownership is ambiguous. The most common edge case is shared infrastructure where one team owns the platform, another owns the application, and a third owns the secrets store; in that model, security must define the policy decision rights up front or reviews become performative. NHIMG’s 52 NHI Breaches Analysis helps illustrate how ownership gaps and lifecycle failures repeatedly turn into exposure. External guidance from the CSA MAESTRO agentic AI threat modeling framework supports the same conclusion for autonomous systems: governance must be shared, but accountability cannot be diffuse.

Related resources from NHI Mgmt Group

• Who should own governance when humans and AI agents share access paths?
• Who should own governance when humans, services, and AI agents all access the same resources?
• How should organizations approach the governance of AI agents?
• Why do AI agents and service accounts create the same governance problem?