TL;DR: Unosecur says many enterprises can count employees but not AI agents, because agents are created inside SaaS, cloud, and development environments outside normal provisioning and can become shadow identities before security teams notice. Continuous discovery and unified inventory matter because identity governance breaks when agents appear, authenticate, and gain access faster than manual registration can track them.
At a glance
What this is: This article argues that AI agent discovery is a visibility control for enterprise identity, with continuous discovery needed to find agents that appear outside normal provisioning and quickly become shadow identities.
Why it matters: IAM, IGA, and security teams need agent discovery because inventory gaps make it impossible to judge access, ownership, and lifecycle for autonomous or semi-autonomous non-human identities.
👉 Read Unosecur's analysis of AI agent discovery and enterprise visibility gaps
Context
AI agent discovery is the continuous identification of agents across cloud, SaaS, and development environments. In identity terms, it closes an inventory gap: you cannot govern what you cannot see, and agents often appear outside the normal provisioning path used for human identities or traditional service accounts.
The article's governance problem is straightforward. Security teams usually know how many people they have, but agents can be created inside tools, inherit access from connected platforms, and begin acting before any formal registration happens. That makes agent visibility an identity governance issue, not just an asset discovery issue.
The article also ties discovery to lifecycle context. Once an agent is visible, teams still need to understand what models, tools, and knowledge sources sit behind it, because those dependencies define its effective access profile and shape how it should be governed over time.
Key questions
Q: What breaks when organisations pre-provision identities for ephemeral AI agents?
A: Pre-provisioning creates standing accounts for actors that may never recur, which increases credential sprawl and leaves residual access after the task ends. It also makes the environment harder to review because the identity record can outlive the actual operational need.
Q: Why do AI agent inventories need continuous discovery instead of periodic audits?
A: Periodic audits are too slow for identities that can appear, authenticate, and act within the same work session. Continuous discovery captures the signals that matter in real time, which is the only way to keep the inventory aligned with active agent behaviour rather than yesterday's records.
Q: What signals show that an AI agent inventory is actually useful?
A: A useful inventory goes beyond a name and status field. It should show the model behind the agent, the tools it can invoke, and the data sources or knowledge base shaping its context. If those elements are missing, the inventory does not support a meaningful access review.
Q: How should security teams govern AI agents that inherit authority from other identities?
A: Security teams should govern AI agents by tracking identity lineage, not just credentials. That means recording the originating identity, the delegated authority path, and the runtime context for each action. If an agent can inherit permissions from humans, services, or other agents, policy has to evaluate the full chain before access is granted or continued.
Technical breakdown
Why AI agents evade normal identity inventory
Traditional identity inventories assume that subjects are enrolled through a predictable workflow and then reviewed later. AI agents break that model because they can be created inside SaaS applications, cloud projects, or developer experiments without passing through central provisioning. The result is not merely poor recordkeeping. It is a visibility failure caused by the control plane and the identity plane living in different places. Manual registration and periodic audits cannot reliably catch identities that appear and act between review cycles. For IAM and IGA teams, the technical issue is that discovery must observe creation and authentication signals, not just rely on self-service records.
Practical implication: Instrument identity discovery against creation and authentication events, not only against registration queues.
What continuous discovery changes for AI agent governance
Continuous discovery treats each authentication, tool call, or resource touch as evidence that an agent exists and should remain on the inventory. That matters because agents do not declare themselves once and stay fixed; they can reappear as their runtime context changes. A quarterly review is too coarse for identities that can be spun up in minutes and used immediately. The architectural shift is from static inventory to recurring observation, with discovery acting as a control that keeps pace with non-human runtime behaviour. That is the only way to narrow the gap between known identities and active identities.
Practical implication: Move from point-in-time inventory to continuous signals that confirm agent existence as it operates.
Why model, tool, and knowledge-base context matters
An AI agent's risk is not defined by its label alone. The effective access profile comes from the model it runs on, the tools it can invoke, and the data or knowledge base that shapes its context. Two agents can look similar in a dashboard while having radically different blast radius. One may be read-only and tightly bounded, while another can write to production through a connected tool. Discovery therefore has to surface attached capabilities, not just the existence of the agent. Without that context, teams are cataloguing names instead of governing authority.
Practical implication: Require inventory records to include the agent's model, tools, and data sources before approving governance decisions.
Breaches seen in the wild
- SalesBleed Salesforce Agentforce 2026: Three fixed Agentforce flaws let poisoned web leads make AI agents leak CRM data with zero clicks and send phishing under the agent's identity.
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
AI agent discovery is now a governance prerequisite, not a convenience feature. The core failure in this article is not lack of reporting, but lack of authoritative inventory over non-human actors that are created outside standard identity workflows. Once an agent can appear inside SaaS or cloud and begin acting before registration, the governance model shifts from approval-led to detection-led. That is a structural change for IAM and IGA teams, not a dashboard enhancement.
Shadow AI agents expose a discovery-first control gap. The article describes active, credentialed identities that remain invisible because they bypass the provisioning queue entirely. That means ownership, review, and offboarding all start from an incomplete baseline. Shadow AI agents: identities that are live before they are governed create a durable audit problem, because the inventory itself is already stale when review begins. Practitioners should treat discovery as the entry point to every downstream control.
Inventory without context is not governance. A record that names an agent but omits its model, tools, and knowledge sources still leaves teams unable to assess what the identity can actually do. This is especially important when agents inherit access from the platforms they connect to rather than from a single explicit grant. The result is a hidden capability layer that must be visible before access can be judged as reasonable.
The identity control problem is now cross-domain. The article correctly places AI agents alongside human identities and other non-human identities because the decision question is the same: does this subject belong in the environment, and what can it reach? That convergence matters for unified identity governance, because separate views for people, service accounts, and agents create blind spots that attackers and careless adopters can both exploit. Practitioners need one inventory model across all actor types.
Discovery must be lifecycle-aware or it will decay quickly. An agent discovered today can change exposure tomorrow as it gains tools, loses tools, or is retired and replaced. Governance therefore cannot stop at first sighting. The field needs continuous linkage between discovery, ownership, and retirement so that the inventory stays aligned with the agent's real operating state, not its original registration state.
From our research library:
- Only 5.7% of organisations have full visibility into their service accounts, according to the Ultimate Guide to NHIs.
- 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: Shadow AI and AI Agent Discovery Guide
What this signals
Discovery is the control that makes AI agents governable. Once agents can be created outside normal provisioning, the main failure mode is not permission design but absence of inventory. Security teams need continuous discovery across cloud, SaaS, and development environments so the control surface starts with presence, not assumption.
Unified identity views matter because AI agents do not exist in isolation. An agent's effective authority depends on what sits around it: the human users, service accounts, and machine identities it touches. Separating those views turns access evaluation into manual correlation, which is exactly how shadow identity risk persists.
For practitioners
- Implement continuous AI agent discovery Scan cloud, SaaS, and development environments continuously so agents are surfaced as soon as they appear, not at the next quarterly review.
- Unify agent inventories with broader identity data Place agents in the same operational view as human identities and other non-human identities so ownership and access can be evaluated together.
- Record agent runtime context Capture the model, tools, and knowledge base connected to each agent so reviewers can assess effective authority, not just name and status.
- Tie discovery to lifecycle governance Require discovery output to feed ownership, review, and decommissioning processes so an agent does not remain active after its purpose changes.
Key takeaways
- AI agent discovery addresses a basic governance problem: organisations cannot manage identities they do not know exist.
- The article's central warning is that agents are created outside normal provisioning and can become shadow identities before anyone reviews them.
- Continuous discovery, unified identity views, and runtime context are the controls that turn agent inventory into usable governance.
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 OWASP Agentic AI Top 10 address the attack and risk surface, while NIST CSF 2.0 and NIST Zero Trust (SP 800-207) set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-10 — Human Use of NHI | Agents created outside normal workflows become hidden identities that people must still govern. |
| NHI-01 — Improper Offboarding | Discovery and lifecycle continuity are necessary because agents can remain active after their purpose changes. | |
| Recommendation — Map unmanaged agent creation to NHI-10 and require discovery before any access review. Tie discovery outputs to offboarding so retired agents are removed from inventory and access paths. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | The article is about whether agent access can be seen and evaluated at all. |
| Recommendation — Use PR.AA-05 to keep agent entitlements visible, reviewable, and aligned to actual use. | ||
| NIST Zero Trust (SP 800-207) | Continuous verification | Continuous discovery mirrors zero trust's requirement to verify identities as they operate. |
| Recommendation — Apply continuous verification to confirm agent identity presence before trust is extended. | ||
| OWASP Agentic AI Top 10 | ASI03 — Identity & Privilege Abuse | Unseen agents can inherit and use privilege without governance, which is the core risk described. |
| Recommendation — Track agent privilege assignment to ASI03 and constrain tools before broad access accumulates. | ||
Key terms
- AI Agent Discovery: AI Agent Discovery is the process of finding and cataloging software agents that operate with some independence across systems, data, and workflows. It identifies where agents run, what they can access, which identities they use, and how they connect to tools, APIs, and other services for governance and security.
- Shadow AI Agent: A Shadow AI Agent is an AI-driven software entity that operates outside approved governance, visibility, or control. It may access data, call tools, or make decisions without formal registration, policy enforcement, or security review. In practice, it creates hidden identity, access, data, and audit risk across enterprise environments.
- Unified Identity Visibility: Unified identity visibility is the ability to map human and non-human access across data and AI environments in one view. It connects identities, entitlements, and activity so teams can see who or what is touching sensitive data. This visibility is the foundation for least privilege, monitoring, and governance.
- Agent Runtime: The agent runtime is the execution environment where an AI agent reads data, calls tools, and carries out actions. It matters because the runtime is where identity, policy, and filesystem boundaries either hold or fail. If those boundaries are weak, the agent becomes a high-privilege path into the environment.
What's in the full article
Unosecur's full article covers the operational detail this post intentionally leaves for the source:
- How Discover Agents surfaces AI agents across cloud, SaaS, and development environments.
- How the unified identity view correlates agents with human identities and other non-human identities.
- How underlying models, tools, and knowledge bases are mapped for each agent.
- How the platform frames lifecycle continuity from provisioning through retirement.
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 or identity governance programme, it is worth exploring.
Published by the NHIMG editorial team on September 6, 2026.
Updated on October 8, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org