TL;DR: AI-driven identity operations are shifting from quarterly human certification to continuous agent-led decisions, according to Lumos, as the article argues that attackers increasingly target legitimate identities rather than traditional perimeter controls. The governance problem is no longer review volume alone, but whether IAM can keep pace with machine-speed access change.
At a glance
What this is: This is Lumos’s argument that identity governance for AI agents and other non-human identities is shifting from periodic human review to continuous agent-led enforcement.
Why it matters: It matters because IAM teams now have to govern access that changes faster than human review cycles, especially where AI agents, service accounts, and standing privileges intersect.
Context
The core problem is identity governance that still depends on human-paced review cycles while access decisions now change at machine speed. In practice, that means privileged access can be granted, abused, or left standing long before a quarterly certification ever reaches it.
For AI agents, the governance gap is sharper because the actor itself can operate continuously, at scale, and across both human and non-human identity estates. The article frames this as a shift from static administration to continuous access control across service accounts, API keys, and deployed AI agents.
Lumos positions the answer as an always-on fleet of specialised identity agents rather than a larger manual team. The underlying issue is not review volume alone, but whether identity governance can keep up with the tempo of modern access change.
Key questions
Q: What breaks when AI agent access is reviewed only after the fact?
A: After-the-fact review leaves a gap between action and containment. If an agent can already reach a dataset, API, or SaaS system, the damage may be done before a human sees the alert. Runtime checks reduce that gap by stopping unauthorized actions before they execute.
Q: Why do standing API keys and service-account secrets increase compromise risk?
A: Because they create a durable trust path that attackers or unauthorised workloads can reuse after discovery. When credentials do not expire quickly, the security team has to rely on perfect revocation discipline, and that rarely holds across pipelines, cloud services, and partner integrations.
Q: What are the signs that non-human identity governance is starting to slip?
A: Warning signs include reliance on manual upgrade paths, ad hoc secret handling, and fragmented access management across tools and environments. The article mentions caveats, dispatch limits, SSO support, and automated update channels, all of which point to the need for controlled operational discipline. When teams cannot explain how identities are provisioned, updated, and constrained, governance is already weaker than it should be.
Q: How should security teams split work between humans and identity automation?
A: Humans should define policy, approve exceptions, and decide on ambiguous cases. Automation should handle repeated access decisions, continuous monitoring, and routine enforcement at machine speed. That split works only if the underlying identity data is accurate enough for machines to trust it.
How it works in practice
Why quarterly access reviews break down for AI agents
Traditional access review assumes a human reviewer can inspect permission sets, decide what still matters, and certify or revoke access before risk accumulates. That model weakens when AI agents and service accounts change state continuously, because entitlements can appear, expand, or become stale faster than a person can process them. In large estates, the problem is not only scale but timing: access decisions happen in a live operational loop, not a scheduled governance cycle. The result is that certification becomes an after-the-fact record-keeping exercise instead of a control point.
Practical implication: Treat periodic review as a governance backstop, not the primary control for high-change AI and NHI access.
Standing keys and over-scoped non-human identities
A standing key is a long-lived credential that keeps working until someone revokes it. When that key belongs to a service account or workload, the real risk is persistence without ownership, because the credential can outlive the workflow, the application change, or the team that created it. Over-scoped non-human identities are especially dangerous because they can retain rights that no current business process actually needs. In that condition, attackers do not need to break the system first. They only need to find the privilege that was never retired.
Practical implication: Inventory standing non-human credentials by owner, scope, and business purpose before they become invisible access paths.
How an identity agent force changes governance mechanics
The article describes a model where specialised agents handle routine access decisions continuously and escalate only the exceptions that need judgment. Mechanically, that shifts governance from batch certification to live policy execution, supported by identity context such as ownership, app permissions, and approval paths. The important detail is not that the work is automated, but that it is always operating against a current view of who can do what. That makes identity governance less dependent on periodic sampling and more dependent on accurate, continuously refreshed identity data.
Practical implication: Build governance around live identity context and exception handling instead of relying on queue-based certification workflows.
NHI Mgmt Group analysis
Manual access review is no longer the right control plane for agent-speed identity change. Quarterly certification was designed for relatively slow-moving human access, not for AI agents and service accounts whose privileges can shift continuously. When identity state changes faster than review cadence, the review process becomes observational rather than preventive. The implication is that identity governance must move its control point upstream, to issuance and live enforcement.
Standing non-human access is now a persistence layer for attackers. The article correctly highlights that attackers are increasingly going after legitimate identities rather than the perimeter. That changes the threat model for NHIs because the real exposure is not just credential theft, but long-lived access that remains usable after the original business need has moved on. Practitioners should read this as a lifecycle failure, not a tooling gap.
Identity blast radius is the new governance unit. The most important question is no longer how many identities exist, but how much damage a compromised identity can still do. Service accounts, API keys, and AI agents all widen blast radius when ownership, scope, and purpose are unclear. The right governance response is to make exposure measurable at the identity level rather than only at the application level.
Access governance for AI agents collapses the assumption that reviewable access persists long enough to be reviewed. That assumption was designed for human-paced certification cycles. It fails when an autonomous or semi-autonomous system can acquire, use, and discard access inside an operational loop without waiting for a human gate. The implication is not merely more automation, but a different governance model for time, ownership, and accountability.
Identity teams are becoming policy engineers, not ticket processors. The article points toward a durable shift in operating model: humans should define policy boundaries, while machine processes execute the routine decisions. That is a useful direction only if the underlying identity context is trustworthy and current. Practitioners should focus on where policy is expressed, where exceptions are judged, and how quickly those judgments become enforceable rules.
From our research library:
- 69% of security leaders agree identity management must fundamentally shift to address agentic AI systems, according to the 2026 Infrastructure Identity Survey.
- 19% of organisations give AI systems dramatically more access than human employees, nearly one in five granting unrestricted privilege, according to the 2026 Infrastructure Identity Survey.
- Read next: AI Agent Authorisation Guide
What this signals
Identity governance for AI agents is becoming a live-control problem, not a periodic-review problem. Access reviews were built for slower human workflows, but AI agents and other non-human identities can change scope continuously. That means practitioners need governance logic that can act at issuance time, not just at the next certification window.
Identity blast radius is the concept teams should start measuring. When service accounts, API keys, and AI agents all hold production rights, the question is how much damage any one identity can do if it is misused. The safest programmes reduce standing scope first, then automate only where the data is current enough to trust.
69% of security leaders agree identity management must fundamentally shift to address agentic AI systems, according to the 2026 Infrastructure Identity Surveyaccording to the 2026 Infrastructure Identity Survey. That tracks with the operational reality: once access changes faster than humans can certify it, governance has to move from review cadence to continuous enforcement.
For practitioners
- Map every standing non-human credential Identify service accounts, API keys, and other non-human identities that still have persistent access, then tie each one to an owner, workload, and business purpose. Remove credentials that cannot be linked to an active use case.
- Move routine access decisions out of quarterly review Reserve human review for exceptions and high-risk grants, while using continuously refreshed policy to handle low-risk, repetitive access decisions that change too quickly for manual certification.
- Reduce privilege on AI and workload identities Scope AI agents, service accounts, and workload identities to the minimum access needed for their current function, then validate that their permissions still match the workflow they actually run.
- Track identity context before automation scales Maintain current ownership, approval paths, and app-level permission data so automated governance can make accurate decisions instead of amplifying stale records.
- Escalate only genuine edge cases to humans Design the operating model so people handle exceptions, anomalous grants, and policy disputes, while routine access is enforced by machine policy in the background.
Key takeaways
- AI agent governance is pushing identity teams away from batch certification and toward continuous control over access that changes too quickly for quarterly review.
- The main risk is not only more identities, but more standing access without clear ownership or current business justification.
- Practitioners should focus on live identity context, exception handling, and least-privilege scope before automation scales further.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
OWASP Non-Human Identity Top 10 and MITRE ATT&CK address the attack and risk surface, while NIST CSF 2.0 sets the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-05 — Overprivileged NHI | The article focuses on service accounts and AI agents with excess standing access. |
| NHI-07 — Long-Lived Secrets | Standing keys and persistent credentials are central to the article's governance problem. | |
| NHI-01 — Improper Offboarding | The article highlights access that remains after ownership or business need has moved on. | |
| Recommendation — Apply least-privilege scoping to non-human identities before access becomes operationally persistent. Reduce credential lifetime and revoke long-lived secrets that outlast the workflow they support. Offboard non-human identities when the workflow changes and remove access that no longer has a current owner. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | The article is about governing entitlements continuously across human and non-human identities. |
| Recommendation — Continuously validate entitlements so access decisions reflect current business need, not stale certification records. | ||
| MITRE ATT&CK | TA0006;TA0008 — Credential Access; Lateral Movement | The article describes attackers targeting legitimate identities and abusing them at scale. |
| Recommendation — Map identity exposure to credential access and lateral movement paths in your detection and response plans. | ||
Key terms
- Identity Agent: An identity agent is software that helps enforce or orchestrate identity controls across users, systems, and automated workloads. In an agentic environment, it may support access decisions, policy application, or lifecycle actions. The control value depends on governance, because automation without oversight can amplify access risk.
- Standing Access: Standing access is persistent privilege that remains available without fresh approval or contextual checks. In NHI environments, standing access usually appears as long-lived tokens, reusable service accounts, or broad roles attached to automation. It is convenient operationally, but it expands risk when conditions change or secrets leak.
- Access Certification: Access certification is the periodic review of whether an identity still needs its current entitlements. For NHIs, certification is only reliable when reviewers know the identity's owner, purpose, and expiry, otherwise stale machine access can persist long after the original use case has ended.
- Identity context: The entitlement, ownership, and purpose information that explains why an action occurred and whether it was expected. For security operations, identity context turns raw alerts into decisions by showing which human or non-human identity acted and what it was allowed to do.
Deepen your knowledge
NHI governance, agentic AI identity, and machine identity security 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.
Published by the NHIMG editorial team on June 20, 2026.
Updated on October 10, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org