Because review cycles assume privilege persists long enough to be observed, challenged, and certified. Machine and agent access can be created by automation, used quickly, and then hidden behind the next workflow step, so certification alone misses the period when the access was actually active.
Why access reviews miss non-human and agent access
Access reviews are usually designed around stable, human-held entitlements that can be checked on a schedule. Non-human access often behaves differently: it is provisioned by automation, consumed by services or agents, and then changed or discarded before the next certification cycle. That mismatch means the review can be formally complete while still failing to reflect actual runtime exposure.
Where the review model breaks down
The core problem is timing. Review campaigns ask whether access should still exist, but machine and agent access is often created, delegated, and used inside a much shorter window than the review period. If the control only captures a monthly or quarterly snapshot, it may never see the access in the state that mattered operationally.
Reviews also depend on a clear, accountable reviewer and a meaningful context for judgment. That is harder when the entitlement belongs to an application, workload, integration, or autonomous agent rather than a person. In practice, reviewers may see a service account name, an API token, or a generic agent label without enough business context to decide whether the access is necessary, risky, or already stale.
The result is often rubber-stamped certification campaigns that validate ownership on paper but do not verify whether the access path was bounded, monitored, and removed in time. That is why identity governance has to treat access review as one control in a broader lifecycle, not the control that makes non-human access safe by itself.
Why automation and agentic access are especially hard to certify
Automation changes access faster than review workflows do. A pipeline may create credentials, run a task, and rotate or retire them in the same day. An agent may request tool access dynamically, use it for a narrow objective, and then move on to another task or context. By the time an approver opens the review item, the access may already have been replaced, expired, or hidden behind another workflow layer.
That makes certification a weak detector for short-lived or event-driven privilege. It also makes static entitlements misleading, because the real control question is not only who owns the account, but when it can act, what it can reach, and whether the action is still observable after the task completes. Good governance therefore needs lifecycle evidence, not just a periodic attestation.
NHI lifecycle management helps because it focuses on provisioning, rotation, offboarding, and visibility across the full lifecycle. For the same reason, Joiner-Mover-Leaver processes matter when tokens, keys, and agents can outlive the human event that created them.
Risk and Threat Considerations
When reviews miss non-human access, the risk is not just audit inaccuracy. Excess privilege can persist long enough for a service, script, or agent to be abused, and the access may be hard to distinguish from legitimate automation because it looks like normal system activity. That creates a detection gap as well as a governance gap.
Failure mechanism: Review cycles certify a snapshot, while the risky access exists in a shorter, more dynamic window. If lifecycle events, ownership changes, or credential rotation are not tied to review evidence, the control can repeatedly approve stale or already-exercised access.
Impact: Privilege can remain active without effective challenge, enabling unauthorized use, lateral movement, or silent overreach by a machine workload or agent. Over time, this weakens least privilege and makes access governance look stronger than it is.
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, OWASP Agentic AI Top 10, MITRE ATT&CK and OWASP API Security Top 10 address the attack and risk surface, while NIST SP 800-53 Rev 5 sets the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-01 — Improper Offboarding | Missed review windows leave non-human access active after it should have been removed. |
| NHI-07 — Long-Lived Secrets | Short review cycles miss secrets that remain usable far beyond the access window. | |
| NHI-05 — Overprivileged NHI | Access reviews often fail to catch excessive machine or agent privilege before abuse. | |
| Recommendation — Tie reviews to offboarding and revoke expired non-human access automatically. Replace long-lived secrets with rotated, expiring credentials and monitor their usage. Scope non-human entitlements to least privilege and recertify high-risk access more frequently. | ||
| NIST SP 800-53 Rev 5 | IA-5 — Authenticator Management | Credential lifecycle and rotation are central when access changes faster than reviews. |
| AC-2 — Account Management | Reviews depend on accurate account inventory, ownership, and timely removal of dormant access. | |
| Recommendation — Manage non-human credentials with expiration, rotation, and revocation controls. Maintain an authoritative inventory of non-human accounts and remove stale access promptly. | ||
| OWASP Agentic AI Top 10 | ASI03 — Identity & Privilege Abuse | Agents can hold and use access in ways periodic certification does not capture. |
| Recommendation — Constrain agent permissions and validate runtime authorization before each tool use. | ||
| MITRE ATT&CK | T1078 — Valid Accounts | Abuse of legitimate machine credentials is a common path when reviews miss active access. |
| T1098 — Account Manipulation | Automation can create or modify access between review periods, defeating snapshot-based checks. | |
| Recommendation — Hunt for legitimate-account abuse when non-human access persists beyond its intended window. Monitor for automated account and entitlement changes that occur outside review cadence. | ||
| OWASP API Security Top 10 | API2 — Broken Authentication | Machine and agent access commonly depends on credentials that outlive the review cycle. |
| API5 — Broken Function Level Authorization | Broad function access can remain hidden when reviews focus on account presence, not actions. | |
| Recommendation — Validate how non-human clients authenticate and rotate credentials before they become stale. Check that non-human callers can invoke only the functions they are meant to use. | ||
Practitioner Guidance
What to verify: Confirm that every non-human entitlement has a named owner, a defined purpose, an expiry or rotation signal, and an event trail that shows when it was actually used. If the review record cannot answer those four questions, the certification is not trustworthy enough to rely on alone.
Decision rule: If access can be created and consumed faster than the review cycle, shift the control emphasis to lifecycle enforcement, automated expiry, and exception handling, then use the review as a backstop rather than the primary safeguard. If the access is long-lived and high privilege, require stronger reviewer context and tighter evidence before recertification.
Common mistake: Treating service accounts, API credentials, and agents as if they were low-risk because no human sits behind them. The operational reality is the opposite: they are often the fastest path to production impact, so they need stronger inventory, context, and retirement discipline than ordinary user access.
Practitioner takeaway: Access review only works for non-human access when it is paired with lifecycle control and usage visibility, because certification without timing evidence is easily outpaced by automation.
Related resources from NHI Mgmt Group
- How should security teams run access reviews for non-human identities?
- How should security teams govern non-human identities that have persistent access?
- Why do periodic access reviews often fail to control entitlement drift in large organisations?
- Why do manual access reviews often fail to keep Microsoft 365 permissions under control?