Who should own the decision to self-host agent governance infrastructure?

Identity, security architecture, and platform teams should decide it together, because the answer affects IAM policy, network egress, secrets handling, and audit design. If one team owns only the tool and another owns the boundary, accountability will break during review or incident response.

Why This Matters for Security Teams

The ownership question is really a governance question about where accountability lives when agent controls touch identity, network boundaries, logging, and secrets. Self-hosting can reduce vendor exposure and improve inspection depth, but it also shifts responsibility for policy enforcement, runtime isolation, and evidence collection into the enterprise. That makes “who decides” as important as “what stack to deploy.” Current guidance from NIST AI Risk Management Framework and the OWASP Agentic AI Top 10 points to shared accountability across governance and operations, not a single tool owner.

NHIMG’s The 2026 Infrastructure Identity Survey found that only 44% of organisations have implemented any policies to manage AI agents, even though 92% say governance is critical, which shows how quickly tooling decisions outrun operating model decisions. In practice, many security teams encounter ownership gaps only after a review, incident, or policy exception has already exposed the boundary mismatch.

How It Works in Practice

The best operating model is a joint decision with clear decision rights: identity owns credential shape and lifecycle, security architecture owns trust boundaries and control objectives, and platform engineering owns implementation, observability, and day-two operations. That division matters because self-hosted governance infrastructure usually sits in the path of agent authentication, policy checks, approval workflows, and audit logging. If any one of those is owned in isolation, the control plane becomes brittle.

Practitioners often use a simple rule set:

• Identity teams define how agents authenticate, rotate secrets, and prove workload identity.

• Security architecture defines where policy must be enforced, what evidence is retained, and how exceptions are handled.

• Platform teams run the service, patch dependencies, manage availability, and integrate it into CI/CD and runtime environments.

This model aligns with the CSA MAESTRO agentic AI threat modelling framework, which treats agent governance as a system of control points rather than a single product. It also fits the reality described in NHIMG’s OWASP NHI Top 10, where identity, secrets, and tool access failures often cascade across teams. Self-hosting is strongest when the organisation can enforce policy-as-code, short-lived credentials, and auditable runtime decisions at the same layer.

That approach also creates a cleaner response path during incident handling: the team that approves the architecture can trace the blast radius, the team that runs the platform can revoke access, and the team that owns identity can prove who or what was authorized. These controls tend to break down when the environment is a fast-moving multi-cloud build with no single owner for runtime policy, because each team sees only part of the control chain.

Common Variations and Edge Cases

Tighter ownership boundaries often increase coordination cost, so organisations must balance speed against the risk of fragmented accountability. There is no universal standard for this yet, especially for smaller teams that want self-hosted governance without building a full internal platform.

In mature environments, a central platform team may operate the stack while security architecture retains approval authority for policy changes and identity retains ownership of agent credentials. In smaller organisations, the same person or team may wear two hats, but the decision rights still need to be explicit. This is especially important when agent workflows span multiple toolchains, because one team may control the host, another the policy engine, and a third the secrets vault.

The tradeoff becomes sharper for regulated workloads and for agents that can take actions across production systems. Guidance from the NIST Cybersecurity Framework 2.0 and the 2024 ESG Report: Managing Non-Human Identities reinforces that governance maturity is measured by accountable control ownership, not by where the software is installed. In edge cases such as hybrid cloud, mergers, or outsourced platform operations, the operating model should be documented before deployment, or self-hosting will simply relocate ambiguity instead of reducing it.

Related resources from NHI Mgmt Group

• Why is single-provider AI agent governance not enough for enterprise security?
• Who should own the decision to block a release after a safety regression?
• What makes agentic AI an NHI governance issue?
• What is the difference between attack surface management and NHI governance?