Omada Agent Governance shows the identity gap for AI agents

At a glance

What this is: Omada’s announcement argues that AI agent governance needs the same identity discipline already used for people, because existing tools do not provide enough visibility, accountability, or access oversight for digital actors.

Why it matters: IAM, IGA, and PAM teams need a model for AI agents before unmanaged access, orphaned identities, and over-privilege become normal across human and machine identity programmes.

By the numbers:

• 72% of organisations have experienced or suspect they have experienced a breach of non-human identities, 46% confirmed and 26% suspected.
• 70% of organisations grant AI systems more access than they would give a human employee performing the exact same job.
• Only 44% of organisations have implemented any policies to manage their AI agents, despite 92% agreeing that governing AI agents is critical to enterprise security.

👉 Read Omada Identity's announcement on AI agent governance and identity control

Context

AI agent governance is the discipline of defining ownership, access, and oversight for software entities that can act with increasing independence inside enterprise systems. Omada’s announcement lands in a market where organisations are expanding the number of non-human identities faster than they are building the controls to govern them.

The problem is bigger than discovery. Once AI agents can connect to systems, read data, and execute tasks, IAM and IGA teams need to answer who owns the identity, what it can reach, and whether that access still matches the work being performed. That is why the topic belongs alongside non-human identity lifecycle management, not as a side issue in automation planning.

Key questions

Q: How should security teams govern AI agents as identities?

A: Security teams should govern AI agents as identities by assigning ownership, defining purpose, scoping access, and placing them into lifecycle review. The key is to treat the agent like a managed non-human identity, not a generic automation asset. That means access approval, recertification, and retirement decisions must be tied to business accountability.

Q: Why do AI agents complicate identity governance?

A: AI agents complicate identity governance because they can change activity patterns faster than traditional review cycles can observe. They may also persist beyond the original business need, which makes ownership and entitlement validation harder. The result is a governance gap between what the organisation thinks the agent should do and what it can actually reach.

Q: What breaks when AI agents are managed like ordinary service accounts?

A: What breaks is accountability. Service account thinking usually assumes a stable function, a stable owner, and predictable usage. AI agents can be more dynamic, more context-sensitive, and more widely connected, so the same controls may miss orphaned access, inflated permissions, and unclear operational responsibility.

Q: Who should own AI agent access decisions?

A: AI agent access decisions should sit with the business and identity owners who can explain why the agent exists and what work it performs. Platform teams can operate the controls, but they should not be the only group deciding access. Clear ownership is what turns access management into governance rather than simple administration.

How it works in practice

AI agent governance and non-human identity scope

AI agents behave like non-human identities because they authenticate to systems, consume permissions, and operate inside enterprise workflows, but they are not the same as service accounts. An AI agent can change behaviour based on context, which means static entitlement thinking breaks down quickly. Governance has to capture identity, dependency, and action history, not just provisioning state. The practical issue is that access may be technically valid while still being operationally inappropriate for the task the agent is performing.

Practical implication: map AI agents into identity governance processes with explicit ownership and access review triggers.

Why accountability breaks down for AI agent identities

Traditional IGA assumes a stable owner, a stable purpose, and a reviewable access model. AI agents complicate all three because they may be spun up dynamically, copied across environments, or left running after the original business need has changed. That creates orphaned or ambiguous ownership, especially when the agent is treated as infrastructure rather than an identity subject. Without accountability metadata, governance evidence becomes incomplete even when access technically exists in a vault, directory, or platform.

Practical implication: require named ownership, business purpose, and retirement criteria for every AI agent identity.

Access versus usage in agent governance

The announcement highlights a useful distinction between what an agent can access and what it actually uses. That matters because over-provisioning often persists when teams equate broad permission with convenience or future flexibility. A governance model for agents should compare granted access against observed usage to identify excess privilege, stale entitlements, and cross-system reach that no one can justify. This is where IGA, PAM, and workload identity controls overlap instead of operating as separate silos.

Practical implication: review actual agent usage against assigned permissions and remove unnecessary reach.

NHI Mgmt Group analysis

AI agent governance is now an identity problem, not only an AI problem. Once an agent can access systems, move data, and execute tasks, it becomes a governed identity subject rather than a simple automation asset. That shifts the control question from model quality to ownership, entitlements, and evidence. Practitioners should treat AI agent sprawl as part of the broader non-human identity population.

Accountability breaks when agent identities outlive the business purpose that created them. Orphaned AI agents are the same governance failure pattern as abandoned service accounts, but with faster creation and wider reach. If ownership, purpose, and retirement are not recorded, audit evidence becomes partial and risk decisions become guesswork. Practitioners should align AI agent lifecycle control with existing IGA discipline.

Identity governance for AI agents will increasingly be measured by least-privilege alignment, not by the number of identities discovered. Discovery tells you scale, but it does not tell you whether access is justified. The stronger signal is how often agent permissions exceed actual task need, especially across cloud and data platforms. Practitioners should expect over-privilege to remain the primary governance gap until access is continuously compared with usage.

Omada’s move reflects a broader market shift toward treating AI agents as first-class governed identities. That direction aligns identity security, compliance evidence, and operational accountability under one programme rather than separate tool chains. The implication for practitioners is that AI agent governance will sit alongside NHI, IGA, and PAM, not outside them.

From our research:

• 70% of organisations grant AI systems more access than they would give a human employee performing the exact same job, according to the 2026 Infrastructure Identity Survey.
• Only 13% of organisations feel extremely prepared for the reality of agentic AI, which helps explain why governance is lagging adoption.
• Ultimate Guide to NHIs is the next step for teams defining ownership, lifecycle, and access controls for machine identities.

What this signals

Over-privilege is becoming the default AI governance failure mode. With 70% of organisations already granting AI systems more access than a human would receive for the same job, according to the 2026 Infrastructure Identity Survey, the gap is no longer theoretical. Teams should expect access governance to become a primary control point for AI agent programmes, especially where cloud, data, and workflow permissions overlap.

Identity programmes need a named concept for agent sprawl. We call it AI agent privilege drift: the steady widening of access, ownership ambiguity, and review debt as agents are copied, extended, and left running. In practice, this means discovery alone is insufficient unless it is paired with lifecycle controls and explicit retirement criteria.

As AI agent adoption grows, governance teams should prepare for review fatigue if they continue using human-centric attestation models unchanged. The better operating model is to anchor AI agent oversight in purpose, usage, and accountable ownership, then map those controls into NIST Cybersecurity Framework 2.0 and identity lifecycle practices.

For practitioners

• Inventory AI agents as governed identities Create a register that records owner, business purpose, system reach, and retirement condition for every AI agent. Keep the register aligned with identity lifecycle reviews so orphaned agents do not disappear into infrastructure inventories.
• Compare granted access with actual agent usage Review permissions against telemetry from data platforms, cloud services, and execution logs to identify excess reach. Prioritise agents with broad access but narrow operational use, because that is where over-privilege hides.
• Extend access review and recertification to AI agents Fold AI agents into recurring governance reviews with the same expectation of evidence, attestation, and removal decisions that apply to other non-human identities. Tie the review to business ownership rather than platform ownership.
• Separate discovery from control decisions Use discovery data to find AI agents, but base governance decisions on ownership, purpose, dependencies, and required access. Discovery alone does not tell you whether an agent should keep its current permissions.
• Align identity controls with audit evidence needs Document how access, usage, and approval records will be produced for audits before the next AI agent rollout. Evidence must show why the identity exists, what it can reach, and who accepted the risk.

Key takeaways

• AI agent governance fails when organisations treat autonomous software as a simple extension of existing IAM processes rather than as a distinct identity class.
• The scale problem is already visible, with most organisations granting AI systems more access than a comparable human would receive and many admitting they are not ready for agentic AI.
• Teams should shift from discovery-only programmes to ownership, lifecycle, and usage-based access controls that can justify every agent entitlement.

Key terms

• AI Agent Identity: An AI agent identity is the account or credential set that allows an autonomous software actor to authenticate, access tools, and perform tasks inside enterprise systems. In governance terms, it must be managed as a distinct identity subject with ownership, purpose, and review requirements.
• Orphaned Non-Human Identity: An orphaned non-human identity is a service account, token, or agent credential that remains active after its owner, purpose, or business need has disappeared. These identities are dangerous because they often keep working while accountability has vanished, making access harder to justify and easier to abuse.
• Access Recertification: Access recertification is the repeated review of whether a user or machine identity still needs its existing permissions. For AI agents and other non-human identities, the review must consider actual usage, business ownership, and lifecycle state, not just whether the credential still exists.
• Privilege Drift: Privilege drift is the gradual expansion of access beyond what was originally intended or justified. In AI agent environments, it often appears when agents are copied, repurposed, or connected to new systems without re-evaluating whether their permissions still match the work they perform.

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 identity security capability in your organisation, it is worth exploring.

This post draws on content published by Omada Identity: Omada Agent Governance and the governance gap for AI agents. Read the original.