Shadow AI and NHIs are forcing identity governance to change

At a glance

What this is: This is an analysis of Identiverse 2025 themes, arguing that shadow AI and machine identity sprawl are pushing identity governance from passive visibility toward real-time identity intelligence and automated action.

Why it matters: It matters because IAM programmes that still rely on periodic review and human-centric governance will miss unmanaged AI tools, overloaded machine identities, and access changes that now happen faster than manual controls can absorb.

Context

Identity governance is the set of controls used to discover, approve, review, and revoke access across people, machines, and now AI-driven workflows. The article argues that the old model is failing because visibility alone does not stop unmanaged access from accumulating across a growing identity surface.

The core governance gap is that service accounts, API keys, and AI agents are now central to how work gets done, but traditional identity processes were built around human managers, periodic certification, and slower access change cycles. In that environment, shadow AI becomes a blind spot and machine identities become under-governed by design.

Key questions

Q: What breaks when identity governance still relies on manual approvals and rule maintenance at scale?

A: Manual governance increases error, slows access decisions, and raises the chance that users receive inappropriate or delayed access. It also makes role design, certification, and revocation harder to keep current. As environments grow, teams lose operational consistency and spend more time resolving exceptions instead of preventing risky access patterns.

Q: Why does Shadow AI create new risk in application security?

A: Shadow AI creates risk because code can be shaped by unapproved assistants outside normal review and policy controls. Traditional scanning sees the result, but not the context of how it was produced. That weakens provenance, auditability, and the ability to prove that insecure patterns were intercepted early.

Q: How do security teams know behavioral identity intelligence is actually working?

A: It is working when teams see fewer false positives, faster identification of credible anomalies, and better separation between harmless deviation and real threat. Strong programmes also show that correlation across behavior, access, and threat data leads to earlier intervention. If the system only creates more alerts, it is not delivering useful context or improving response quality.

Q: What should organisations do when machine identities already outnumber human identities?

A: Treat that as a governance design change, not just an inventory problem. Rebuild identity controls around lifecycle enforcement, automated renewal and revocation, and continuous drift detection so the programme matches machine speed instead of reporting on it after the fact.

Technical breakdown

Why human-centric identity governance breaks for machine identities

Traditional identity governance assumes access can be assigned to a person, reviewed by a manager, and adjusted on a predictable lifecycle. That model does not fit service accounts, API keys, or AI agents, which often have no natural owner, no meaningful manager approval path, and access that changes faster than review cycles. When the governed subject is a machine identity, lifecycle management has to account for inventory, ownership, scope, and revocation in ways that human IAM processes do not.

Practical implication: Practitioners need governance controls that treat machine identities as first-class subjects, not as exceptions hidden inside human access workflows.

How shadow AI creates ungoverned identity relationships

Shadow AI appears when employees connect AI tools directly to corporate systems without going through sanctioned identity flows. That creates identity relationships outside IT and security visibility, which means policy cannot be enforced where the access is created. The problem is not just discovery failure. It is that the organisation no longer knows which AI tools hold access, what permissions they received, or whether those permissions are still justified.

Practical implication: Teams should treat unmanaged AI connections as identity sprawl and bring them into inventory, approval, and revocation processes before they become persistent access paths.

Why real-time policy enforcement matters more than periodic reviews

Periodic access reviews are too slow for environments where applications, roles, and integrations shift continuously. Real-time policy enforcement ties context, such as role changes or new shadow applications, to immediate access decisions. That matters because the risk is often created at the moment access is granted or expanded, not weeks later when a review cycle runs. In practice, governance has to move closer to issuance and change events rather than depending on after-the-fact certification.

Practical implication: Security teams should prioritise policy engines that can evaluate access changes at the point of change, not only during quarterly or monthly reviews.

Threat narrative

Attacker objective: The objective is to gain or preserve access through unmanaged AI-linked identity paths that sit outside normal governance and cannot be reliably reviewed or revoked.

1. Entry occurs when employees connect unsanctioned AI tools directly to corporate systems, bypassing normal authentication and governance checkpoints.
2. Credential or access scope is then expanded through unmanaged identity relationships, leaving AI tools or connected accounts with permissions that were never centrally approved.
3. Impact emerges when these shadow connections create blind spots for policy enforcement, auditability, and revocation across the identity estate.

Breaches seen in the wild

• Vercel Context.ai OAuth Supply Chain Breach: Shadow AI app Context.ai OAuth integration exposes Vercel customer data via unmanaged third-party token.
• OmniGPT breach claim 2025: A hacker claims to have leaked 34 million OmniGPT AI chat messages holding users' API keys and credentials; OmniGPT has not confirmed it.

Read and download The State of NHI & AI Agent Breach Report 2026, covering 150+ breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Identity intelligence is replacing identity governance as the operating model because visibility without action no longer contains risk. The article captures a real shift in practitioner expectations: organisations want decisions enforced in real time, not just observed in dashboards. That is a programme design change, not a tooling preference. The implication is that identity teams must measure governance by intervention speed, not inventory depth.

Machine identities have become the dominant governance problem because they outnumber human users while receiving far less oversight. Service accounts, API keys, and AI agents sit at the center of modern operations, but they do not fit human approval models cleanly. That makes lifecycle ownership, scope control, and revocation the controlling questions for the category. Practitioners should re-centre NHI governance around accountability and expiration, not just discovery.

Shadow AI introduces an access layer that traditional IAM can see only after the fact. When employees bind unsanctioned AI tools to corporate systems, the organisation inherits identity relationships it did not create or approve. This is less a policy exception than a structural blind spot. The practitioner conclusion is that unmanaged AI access must be governed as identity sprawl, not treated as a separate innovation problem.

Real-time policy enforcement is becoming the decisive control plane for mixed human and non-human access. Periodic reviews still matter, but they are too slow to govern environments where role changes, application sprawl, and machine-to-machine activity happen continuously. The next control boundary is the change event itself. IAM and IGA teams should expect enforcement to move closer to issuance, not remain stranded in retrospective certification.

Identity surface sprawl: the governance burden now spans people, machines, and shadow AI connections at once. This is the right named concept for the shift the article describes. The implication is that identity architecture can no longer be designed as separate human, machine, and AI tracks. Practitioners need one policy view across the full identity surface, or they will keep missing the relationships that create exposure.

From our research library:

• 63% of organisations surveyed lacked AI governance policies to manage AI or prevent shadow AI, according to IBM's 2025 Cost of a Data Breach Report.

What this signals

Identity surface sprawl: security teams should expect the identity boundary to keep expanding as employee-led AI adoption, service accounts, and API-driven workflows collide. The practical response is to govern identity relationships as a living system, not a catalogue of static entitlements.

Programmes that still depend on quarterly review cadences will continue to miss the point where risk is introduced. The stronger control pattern is to evaluate access when the change occurs, then use automation to remove or constrain what no longer matches policy.

For practitioners

• Inventory AI-linked access paths Map every AI tool, integration, and connected account that touches corporate systems, including unsanctioned tools discovered outside IT approval channels.
• Classify machine identities as governed assets Assign ownership, business purpose, and expiry expectations to service accounts, API keys, and AI agent identities so they can be reviewed like other privileged access.
• Shift policy enforcement to change time Trigger access evaluation when roles change, applications are added, or new integrations appear, instead of waiting for periodic certification cycles.
• Require revocation paths for shadow connections Define how unapproved AI-linked access is disabled, audited, and removed when the underlying tool or integration is no longer sanctioned.
• Correlate identity data with operational context Join identity events to role, application, and endpoint signals so access decisions reflect current conditions rather than stale governance records.

Key takeaways

• The article argues that identity governance is moving toward identity intelligence because dashboards alone do not prevent unmanaged access from spreading across machines and AI tools.
• Machine identities and shadow AI are now central governance problems because they create access relationships that traditional human-centric IAM processes handle poorly.
• The practical shift is toward real-time policy enforcement, lifecycle ownership, and revocation paths that operate at the moment access changes.

Key terms

• Identity Intelligence: Identity intelligence is the layer that turns raw identity data into context about risk, usage, and privilege. It helps teams distinguish harmless access from materially risky access by linking identity records, entitlement patterns, and behavioural signals, which is essential when non-human identities scale faster than manual review.
• 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.
• Machine Identity: The digital identity of a machine, device, or workload, such as a server, container, or VM, used to authenticate it within a network. Sometimes used interchangeably with NHI, though NHI is the broader category.
• Real-Time Enforcement: Real-time enforcement means a system evaluates conditions while the transaction or session is still active, rather than after the fact. For governance programmes, this is the difference between preventing a harmful outcome and only reporting it later.

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.