TL;DR: RSA 2025 conversations highlighted a widening gap between AI-driven identity governance aspirations and the reality of machine identity sprawl, with 45:1 machine-to-human ratios and 60%+ shadow IT visibility issues shaping the discussion, according to Zluri. The governance problem is no longer review volume alone; it is that legacy IAM controls were built for human directories, not fast-moving non-human identity estates.
At a glance
What this is: This is Zluri’s RSA 2025 analysis of how AI-driven identity governance is colliding with machine identity sprawl and visibility gaps in modern estates.
Why it matters: It matters because IAM and IGA teams are being asked to govern service accounts, certificates, keys, access tokens, and AI agents with models built for human directories and review cycles.
By the numbers:
- 2025 is the year when machine identities outnumber humans by 45:1.
- Over 60% of IT resources in a typical organisation exist as either unmanaged or shadow IT.
Context
AI-driven identity governance is the use of automation and analytics to discover identities, evaluate access, and trigger remediation. The governance gap appears when those controls are applied to estates dominated by non-human identities, because the underlying assumptions were built around human directories, slower lifecycle change, and review cadences that can keep pace with people rather than machines.
At RSA Conference 2025, Zluri framed the problem as one of scale and visibility rather than policy intent. The article argues that machine identities, including AI agents, service accounts, certificates, keys, and access tokens, have become the larger governance surface, while many organisations still depend on identity providers, HR systems, and static connectors that miss what is actually in use.
Key questions
Q: How should security teams govern machine identities differently from human users?
A: Security teams should govern machine identities with lifecycle, context, and runtime controls, not human approval workflows. That means separate ownership, purpose-based entitlement, expiry, and revocation for service accounts, bots, and agents. Human IAM can inform the model, but machine access needs faster review, tighter scoping, and automated enforcement.
Q: Why do periodic access reviews miss the real NHI risk?
A: Periodic reviews miss the real risk because they measure entitlement state at a point in time, while NHI exposure can change between cycles. A service account, token, or OAuth app may remain valid long after its purpose has shifted. Continuous visibility and action matter more than retrospective certification.
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: Should organisations prioritise discovery or remediation first for NHI sprawl?
A: Discovery should come first, because remediation depends on knowing which identities exist, where they live, and which ones are actually active. Without complete visibility, remediation workflows can remove the wrong access while leaving the highest-risk identities untouched.
Technical breakdown
Why human-directory governance breaks under machine identity sprawl
Human-centric identity governance assumes a bounded population, stable ownership, and reviewable entitlements. Machine identities break those assumptions because they are created faster, used more broadly, and often exist outside HR-linked records. Service accounts, keys, certificates, and access tokens can be provisioned in application layers where traditional IGA visibility is thin. Once that estate grows faster than discovery and certification processes, the control problem is no longer approval volume but inventory completeness and entitlement truth.
Practical implication: treat discovery coverage as a prerequisite control, not a reporting feature.
What real-time activity data adds to access governance
Static entitlement data tells you what was granted, not what is actually being used. Real-time activity data closes that gap by correlating application usage with access rights, so dormant, excessive, or risky permissions can be identified in context. In SaaS-heavy environments, this matters because entitlement state drifts quickly and an identity may retain access long after the business need has changed. Without usage data, review workflows certify stale facts instead of current access behaviour.
Practical implication: pair entitlement reviews with runtime usage evidence before approving continued access.
Why autonomous AI agents need separate governance treatment
Autonomous AI agents are not just another workload. If they can choose actions at runtime, select tools, and execute without human approval gates, they become non-human identities with decision power, not just automation scripts. That changes the governance model because the risk is not only credential exposure, but delegated action that can happen faster than a human can review it. In that condition, identity governance must account for agent behaviour, not only agent ownership.
Practical implication: classify AI agents by runtime autonomy and govern them separately from static service identities.
NHI Mgmt Group analysis
Machine identity sprawl is now the primary governance problem, not a side effect of cloud growth. Zluri’s RSA readout points to a market where non-human identities have outgrown the human-centric controls still used to manage them. That matters because the control plane is being asked to govern a population whose size, volatility, and ownership model no longer resemble the assumptions behind classic IGA. Practitioners should treat machine identity inventory as the first order governance problem.
Visibility-driven identity governance is becoming the new baseline for SaaS estates. Static identity sources and pre-existing connectors cannot represent how access is actually used across modern application stacks. The practical consequence is that access reviews can certify entitlement records while missing the activity reality that determines risk. Teams need governance models that reconcile granted access with observed behaviour, not one or the other.
Autonomous identity behaviour creates an assumption collapse around reviewability. Access review cycles were designed for identities whose privileges persist long enough to be observed, assigned, and certified. That assumption fails when an AI agent can obtain, use, and discard access within a short runtime window. The implication is that governance must stop treating review as the primary control for every non-human identity.
Shadow AI extends the identity problem beyond inventory into unsanctioned execution paths. Zluri’s framing shows that unmanaged AI usage is not only a technology adoption issue but a governance blind spot. When unsanctioned tools and identities emerge faster than policy enforcement, entitlement controls lose the ability to define the real attack surface. Practitioners should read Shadow AI as a discovery and control integrity problem, not a tooling trend.
Closed-loop remediation is becoming central to identity operations because review latency is now a risk factor. The article’s emphasis on automation reflects a wider operating reality: manual review cadence is too slow for fast-changing machine estates. That does not mean automation is the objective in itself. It means remediation must be tightly coupled to discovery and usage evidence so the identity programme can act on current state, not historic approval records.
From our research library:
- Nearly 60% of IT leaders cite restrictive cost and complexity as a weakness of legacy identity governance, according to the 2025 State of Identity Governance Report.
- Read next: NHI Lifecycle Management Guide
What this signals
Machine identity sprawl is turning IGA into a discovery problem. Once service accounts, certificates, keys, and tokens outnumber human users, the first question is no longer who approved access but whether the estate is even visible. Practitioners should expect their governance programme to shift from periodic certification toward continuous inventory and usage reconciliation.
Shadow AI is a governance signal, not just an adoption trend. When unauthorised AI tools enter the environment, they often carry their own identity and access paths. That means teams need to look at discovery, entitlement, and activity data together, or they will miss where new access is being created outside policy.
Autonomous access changes the meaning of review. If an AI agent can act within a session and release access before a human review cycle runs, certification loses its timing advantage. The control objective shifts toward issuance-time guardrails and runtime containment, especially for sensitive SaaS and cloud workflows.
For practitioners
- Map machine identity inventory coverage Measure how completely your IGA stack sees service accounts, certificates, keys, access tokens, and AI agents across SaaS, cloud, and internal applications. Use the gaps to identify where HR-linked or connector-based discovery is missing identities.
- Separate review logic for human and non-human identities Do not run the same certification workflow for employees and non-human identities. Define different evidence requirements, ownership fields, and approval thresholds for machine identities that do not map to HR records.
- Tie access approval to activity evidence Require runtime usage signals before recertifying sensitive entitlements in SaaS-heavy environments. This helps stop stale permissions from surviving simply because they were once approved.
- Classify autonomous AI agents by runtime behaviour Document which AI agents can choose actions, select tools, and execute without human approval gates. Those agents need governance treatment that reflects their decision authority, not just their system owner.
- Automate remediation for high-risk access paths Use closed-loop workflows to remove or reduce access when discovery and activity data disagree. The goal is to shorten the time between finding an issue and closing the exposure.
Key takeaways
- The article argues that machine identity sprawl has overtaken human-centric identity governance assumptions, especially in SaaS-heavy environments.
- The scale problem is clear in the source article, which says machine identities outnumber humans by 45:1 and over 60% of IT resources may sit in shadow IT or unmanaged states.
- The practical response is to pair discovery, activity data, and separate lifecycle controls so non-human identities are governed as a distinct estate.
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-01 — Improper Offboarding | Machine identities that outlive their business need create governance drift in this article. |
| NHI-05 — Overprivileged NHI | The article describes machine identities governed with human-era access assumptions and excessive scope. | |
| NHI-07 — Long-Lived Secrets | Keys and access tokens are central to the sprawl problem discussed in the article. | |
| Recommendation — Apply offboarding controls to revoke dormant service accounts, tokens, and certificates when ownership changes. Review machine entitlements for least privilege and reduce access that exceeds current task scope. Rotate long-lived machine secrets and replace static credentials where possible. | ||
| MITRE ATT&CK | TA0006;TA0008 — Credential Access; Lateral Movement | The article links identity stack weakness to attacker movement through access paths. |
| Recommendation — Map identity-stack exposure to credential access and lateral movement detections in your threat program. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | The core issue is governing entitlements across machine and human identities. |
| Recommendation — Use entitlement governance to validate that access aligns with current business need and usage. | ||
Key terms
- Machine Identity Sprawl: Machine identity sprawl is the uncontrolled growth of non-human identities across teams, platforms, and business processes. It becomes a governance problem when identities are created faster than they can be inventoried, reviewed, rotated, or retired, leaving security teams with incomplete visibility and weak accountability.
- Shadow AI: AI agents, copilots, or connected tools operating without full visibility or governance from security teams. Shadow AI becomes an identity problem when those systems authenticate with unmanaged tokens, service accounts, or OAuth apps that can reach production resources.
- Closed-Loop Remediation: A governance process that does not stop at finding risk. It removes or reduces access, confirms the change in the source systems, and keeps evidence that the risky condition stayed fixed. For NHIs, this is the difference between inventory and actual risk reduction.
- Runtime Usage Evidence: Observed data showing how an identity is actually being used in production rather than how it was originally granted. For machine and autonomous identities, runtime evidence helps distinguish active, justified access from stale or excessive entitlements.
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.
Published by the NHIMG editorial team on June 11, 2026.
Updated on October 8, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org