Non-human identity security protects the accounts and credentials used by software, services and AI agents. It covers ownership, access, credential lifetime and the data an identity can reach. Begin with your own inventory: published NHI security research describes recurring governance gaps, but there is no universal identity-to-employee ratio.
This guide explains what non-human identities are, the risks behind service accounts and OAuth apps, and the controls to compare when evaluating NHI security tools. The practical question is which identities can reach sensitive data, and who can remove access that is no longer needed.
TL;DR
- AI agents rely on accounts, tokens and permissions that need explicit ownership and review.
- Effective access matters: trace roles, grants and credentials through to sensitive resources.
- Maintain governance evidence: owners, approved access, review records and revocation procedures.
- The fix is systematic: Inventory → Lifecycle → Runtime Constraints.
What are non-human identities (NHI)? A definition
Non-human identities are digital identities assigned to software or services rather than individual people. Credentials authenticate those identities; they are not the same thing as the account or principal. Common examples include:
- Service accounts, technical accounts for applications and batch jobs
- API keys and tokens, access credentials for system-to-system interfaces
- Certificates and secrets, machine authentication in CI/CD pipelines, cloud workloads and containers
- OAuth apps and integrations, third-party connections to SaaS platforms
- AI agents, software that uses one or more identities to access tools, data and services
An AI agent may use a new identity or reuse an existing service account, token or delegated user session. The Cloud Security Alliance’s January 2026 survey highlights ownership, governance and credential-lifecycle gaps. Its figures describe survey respondents, not every organization.
How to evaluate NHI security tools
Use a real workflow in a product evaluation. Ask each vendor to show these controls in the systems you operate, and distinguish supported capabilities from roadmap items:
| Control to evaluate | Evidence to request |
|---|---|
| Identity discovery | Accounts and credentials |
| Ownership | Owner and purpose |
| Effective access | Paths to sensitive data |
| Credential lifecycle | Rotation and revocation |
| Remediation | Owner, change and retest |
A secret scanner can identify an exposed credential. An identity inventory can show that an account exists. An NHI security evaluation should also establish what that credential can access, who owns the identity and how access is revoked. The OWASP NHI Top 10 provides a useful risk checklist.
The four core problems of unmanaged machine identities
1. Zombie identities
Service accounts and API keys whose purpose nobody remembers, whose owner left the company, but which remain valid. Every zombie identity is an unguarded side entrance into your infrastructure.
2. Over-privileged agents
A service account may retain permissions from setup or inherit access through another role. If an agent can use it to export CRM records or read production secrets, that access needs a business justification and a defined owner.
3. The confused-deputy problem
An attacker tricks a legitimate, highly privileged agent into acting on their behalf, for example via prompt injection. The agent itself becomes the attack weapon, and no credentials need to be stolen at all.
4. Missing kill switches
Humans can be locked out, sessions terminated. For many NHIs there is no defined way to shut them down immediately and completely in an emergency, no lifecycle boundaries, no runtime constraints, no emergency stop.
NHI governance and compliance evidence
Treat NHI governance as part of access and risk management. Whether specific legal duties apply depends on the organization, sector and use case; an AI agent does not automatically place every company under every regulation. For Germany, start with the BSI’s NIS2 guidance and document the controls relevant to your organization.
Maintain evidence that a reviewer can inspect: the identity owner, approved purpose, effective access, review date and tested revocation procedure. A policy alone does not show whether an agent can still reach customer data.
The 3-step plan: Inventory → Lifecycle → Runtime Constraints
- Inventory, create visibility. Record accounts, credentials, owners, purposes and reachable resources across connected systems. Ask which identities are unowned and which can access production. Use the OWASP NHI risks to structure the review.
- Lifecycle, govern creation and death. Define a binding process: who may create NHIs? With what maximum permissions (least privilege)? When are credentials rotated? And above all: when and how are identities automatically deprovisioned? Zombie identities are born wherever offboarding for machines doesn't exist.
- Runtime constraints, limit and monitor behavior. Use narrowly scoped, short-lived credentials where supported. Monitor sensitive actions and test revocation, including active sessions and cached credentials. Record how long containment actually takes.
Conclusion: identity is the new perimeter, and it's machine-shaped
Start with the identities that can read secrets, change production or export customer data, in the order set out in the seven NHI risks. Record the access path, confirm the owner and remove unnecessary privileges. Repeat the review when an agent, integration or role changes. This turns an inventory into a practical NHI security process.
On Elmoz
#FAQ
What is a non-human identity?
A non-human identity is a digital identity assigned to software or services rather than to a person. Examples are service accounts, API keys and tokens, certificates, OAuth apps and AI agents.
How do AI agents relate to non-human identities?
An AI agent may use a new identity or reuse an existing service account, token or delegated user session. Each of those needs an owner, approved access and a way to revoke it.
What should an NHI security tool show?
Identity discovery, ownership, effective access including paths to sensitive data, credential lifecycle and remediation. Ask vendors to show each control in the systems you actually run.
Next step: choose one AI agent and trace its identity through to production or customer data. Explore Elmoz’s AI agent security platform to see how an access graph supports that investigation.