Who should own AI identity governance in the enterprise?

Why This Matters for Security Teams

AI identity governance is not just an IAM assignment problem. Once an AI agent can call tools, chain tasks, or request data on its own, ownership has to cover the full lifecycle: provisioning, policy enforcement, monitoring, and revocation. That makes this a cross-functional control plane issue, not a ticket routing question. NHI Management Group’s research on the Ultimate Guide to NHIs and the Top 10 NHI Issues shows why fragmented ownership leaves gaps between identity, security, and platform operations.

The practical risk is that each team tends to optimise its own slice. IAM may manage accounts, security may watch alerts, and platform teams may own workloads, but no one owns the joint behaviour of the AI identity in motion. That gap is exactly where standing privileges, stale secrets, and weak revocation processes persist. Current guidance suggests treating AI identity governance as an enterprise control with named business and technical owners, aligned to security policy and operational reality. In practice, many teams discover the ownership gap only after an agent misuses access, rather than through deliberate design.

How It Works in Practice

Effective ownership starts by separating responsibilities without separating accountability. Identity teams usually own the identity lifecycle, credential standards, and joiner-mover-leaver equivalents for agents. Security teams define policy, risk thresholds, detective controls, and incident response triggers. Platform or application teams own the service runtime, integration points, and enforcement hooks where agent access is actually used. That structure matches the way autonomous systems behave: the identity is not static, and the risk changes as the agent moves between tools and contexts.

For AI agents, the control model should favour workload identity, short-lived credentials, and runtime policy checks over fixed access grants. NIST’s Cybersecurity Framework 2.0 is useful for clarifying governance, while the NIST Cyber AI Profile helps translate AI risk into operational controls. In practice, ownership should be documented in a RACI that covers:

• who approves creation of an AI identity, service account, or agent workload identity
• who sets policy for JIT access, secrets rotation, and token TTL
• who reviews anomalous tool use, lateral movement, and escalation paths
• who can revoke access immediately during incident response

That model works best when the organisation treats agent identity as a governed workload primitive, not an application afterthought. It also benefits from the lifecycle and breach lessons documented in NHIMG’s Lifecycle Processes for Managing NHIs and 52 NHI Breaches Analysis. These controls tend to break down when AI identities are created inside product teams without central policy, because revocation, monitoring, and audit evidence become inconsistent across environments.

Common Variations and Edge Cases

Tighter governance often increases operational overhead, so organisations have to balance speed against assurance. That tradeoff is especially visible in fast-moving AI programmes, where teams want to deploy agents quickly but still need clear control boundaries. There is no universal standard for ownership in every enterprise yet, so best practice is evolving toward a federated model with one accountable owner and several contributing teams.

In highly regulated environments, security may need stronger veto authority over new AI identities, while in platform-led organisations the platform team may operate the control plane and security may define policy-as-code guardrails. The right answer depends on where identity is instantiated and where revocation can be enforced fastest. For some mature programmes, governance is also extended to data owners when agents can query sensitive systems. NHIMG’s Regulatory and Audit Perspectives is useful here because auditors will ask who approved the identity, who monitored it, and who can prove it was removed on time.

The key edge case is shadow AI, where teams spin up agents or service identities outside the central process. That is where ownership models fail first, because nobody can answer who approved access or who would notice compromise. In those environments, governance should shift from informal coordination to explicit policy enforcement and mandatory registration.

Related resources from NHI Mgmt Group

• Who should own AI agent governance in an enterprise identity programme?
• Who should own governance for workload identity in AI systems?
• Who should own non-human identity governance in an enterprise?
• Who should own AI governance across identity and data controls?