The control plane shift at re:Invent 2025 reshapes identity governance

At a glance

What this is: This commentary says AWS re:Invent 2025 signalled a control-plane shift in identity, with runtime behaviour across humans, services and AI agents overtaking older IAM assumptions.

Why it matters: IAM, IGA and PAM teams need to reassess governance models when identity is no longer a static directory concept but the control plane for dynamic execution across actor types.

Context

A control-plane shift in identity means access decisions are moving from a supporting layer into the runtime path that orchestrates humans, services and AI agents. The old model assumed identity was mostly about provisioning and authentication, but modern systems increasingly use identity to decide what can act, call tools and continue execution.

Ory's re:Invent commentary frames this as a generational break in security architecture rather than a minor product change. For IAM, IGA and PAM programmes, the issue is whether governance can keep up with runtime behaviour that changes faster than review cycles and entitlement models were designed to handle.

The underlying governance gap is not just scale. It is that static privilege models and fixed workflows presuppose a stable actor and a predictable path of action, which is increasingly false across human, workload and agentic environments.

Key questions

Q: What breaks when identity is treated as an administrative task instead of a control plane?

A: When identity is treated as administration, organisations lose sight of how authority is created, inherited, and extended across systems. That leads to stale permissions, invisible delegation, and weak accountability when incidents happen. The programme may still issue accounts correctly, but it will not govern actor behaviour well enough to limit blast radius.

Q: Why do static privilege models fail in environments with workloads and AI agents?

A: They assume future access can be predicted at provisioning time. Workloads chain calls and AI agents may select actions dynamically, so the privilege that exists at the start of a task may not match the access needed or used later in the same session. That mismatch creates blind spots in approval, review and accountability.

Q: How can IAM teams tell whether access review coverage is actually meaningful?

A: They should measure what was excluded as well as what was approved. A strong completion percentage over a weak denominator hides the systems, roles, and accounts that never entered the campaign, and those blind spots are often where stale access survives longest.

Q: What breaks when human, workload, and agent identities are governed separately?

A: Separate governance creates inconsistent privilege rules, weak traceability, and gaps in accountability across delegation chains. A human may authorise the workflow, a workload may execute it, and an agent may choose the next step, so policy fragmentation quickly becomes a control failure.

Technical breakdown

Why control planes now absorb identity decisions

A control plane is the layer that coordinates policy, permissions and execution across systems. In older architectures, identity authenticated a user and then stepped aside. In newer environments, identity is increasingly used to gate APIs, workload actions and agent behaviour at the point of execution. That makes identity part of the runtime control loop, not just the login boundary. The practical effect is that governance has to move closer to issuance, delegation and policy evaluation, because waiting for periodic review no longer matches how decisions happen in distributed systems.

Practical implication: map where identity decisions are made at runtime, not only where accounts are provisioned.

Why static privileges break in runtime-heavy environments

Static privilege assumes the actor's future needs are known when access is granted. That works tolerably for predictable roles, but it breaks when services chain calls, humans trigger automation, or agents choose actions based on context. The more execution happens through delegated identity and short-lived access paths, the less useful a one-time entitlement model becomes. What matters is not only who has access, but when access exists, for how long, and under what policy conditions it can be exercised.

Practical implication: replace entitlement snapshots with controls that can govern time-bound and context-bound access.

How agentic AI changes identity governance assumptions

Agentic AI adds a new governance pressure because the actor can select actions and timing in response to runtime context. That does not make every AI-enabled system autonomous, but it does mean the identity boundary can shift mid-task in ways classic IAM does not model well. For governance, the key issue is whether the programme still assumes a human operator behind every meaningful access decision. When that assumption fails, review, approval and attribution processes need to account for machine-paced execution paths.

Practical implication: identify where agent-driven execution bypasses human-paced approval and recertification cycles.

NHI Mgmt Group analysis

The control plane shift is an identity governance problem, not just an architecture trend. Once identity becomes the mechanism that decides runtime action across humans, services and AI agents, old IAM boundaries lose explanatory power. The relevant question is no longer who logged in, but which identity decided what could happen next. Practitioners should treat control-plane identity as a governance domain in its own right.

Static privilege is now a weak assumption across mixed actor estates. The model assumes future access can be predicted and certified ahead of time, but runtime systems increasingly assemble access as work unfolds. That is especially true where workloads and agents inherit, request or discard permissions dynamically. The implication is that entitlement models need to describe execution context, not just ownership.

Access review cadences were designed for stable actors and durable privileges. That assumption holds poorly when the access path is short-lived, delegated or machine-paced. In mixed environments, a quarterly review can certify an arrangement that no longer resembles live behaviour by the time it is completed. Practitioners should question any governance process that depends on access persisting long enough to be observed.

Identity governance is converging across human, workload and agentic domains. The article's value is in showing that these are no longer separate programmes with separate assumptions. The same governance language now has to cover human SSO, workload permissions and AI agent delegation without pretending they are equivalent. Teams that keep these domains siloed will miss the runtime relationship between them.

Runtime identity orchestration is becoming the new control surface for the enterprise. The named concept here is the control plane shift: identity is moving from record-keeping to execution governance. That does not reduce IAM's importance, it expands its scope into the path where systems decide, delegate and act. Practitioners should align governance models to that runtime control surface.

From our research library:

• 90% of IT leaders say properly managing NHIs is essential for a successful zero-trust implementation, according to the Ultimate Guide to NHIs.
• Read next: Ultimate Guide to NHIs — Standards

What this signals

Identity teams should expect more governance pressure at the control plane, where policy evaluation and execution now meet. That shifts attention from account inventories toward runtime authorization boundaries, especially in environments where services and agents act faster than review cycles can catch up.

Control-plane identity: the useful mental model is that identity is no longer only a directory record or login factor. It is becoming the runtime mechanism that decides whether work can proceed, which means governance needs to cover delegation, issuance and execution together.

The programme implication is straightforward: if access review is still your primary control, you are probably governing the aftermath rather than the action. Identity governance now has to understand which actor type is making the decision, when that decision occurs and whether a human is still in the loop.

For practitioners

• Inventory runtime identity decision points Map where identity is used to authorise actions at execution time across humans, workloads and AI agents. Prioritise systems where access is evaluated during orchestration, API calls or agent task execution rather than only at login.
• Review privilege models for dynamic execution Check whether roles, scopes and delegated access still assume stable work patterns. Flag environments where permissions are assembled or consumed inside workflows, because those are the first places static entitlements stop describing reality.
• Align access reviews to live behaviour Measure whether recertification and approval cycles can see the identity states that matter during real execution. If they cannot, move governance earlier in the lifecycle so that issuance and delegation carry more weight than periodic review.
• Separate human, workload and agent governance Document which policies are shared and which are actor-specific. Human SSO, workload permissions and agent delegation should not be governed as if they produce the same risk profile, even when they sit in the same platform.

Key takeaways

• The article frames identity as part of the runtime control plane, which changes what IAM, IGA and PAM teams need to govern.
• Static privilege and periodic review are increasingly poor fits for systems that assemble access dynamically during execution.
• Teams should rework governance around runtime decision points, actor type differences and the timing of delegation.

Key terms

• Control Plane Identity: Control plane identity is the identity used by systems that manage, configure, and orchestrate other systems. It governs who or what can change policies, routes, permissions, or infrastructure state. In practice, it includes human admins, automation accounts, and service identities with elevated authority over control functions.
• Runtime Authorisation: Runtime authorisation is the practice of deciding access while a task is in progress, rather than only at provisioning time. It matters for NHIs because credentials and entitlements can change risk mid-session, especially when automation or AI agents interact with sensitive systems.
• Static privilege model: An access model that assumes future needs can be defined up front and reviewed later through roles, entitlements or fixed workflows. It becomes fragile when permissions are created, consumed and discarded inside live tasks.
• Delegated Identity: Delegated identity is when one actor acts on behalf of another with explicit permission and bounded authority. In AI-assisted commerce, it requires clear consent, limited scope, and traceable records so the retailer can distinguish authorised delegation from unauthorised automation.

Deepen your knowledge

NHI governance, agentic AI identity, and machine identity lifecycle are core topics in our NHI Foundation Level course, the industry's only accredited NHI security programme. If you are building or maturing an IAM programme, it is worth exploring.