Runtime identity control for non-human identities in the machine-driven enterprise

At a glance

What this is: NHI Mgmt Group's analysis, drawing on SSH Communications Security's blog post: PAM must shift from human-centric access control to runtime identity control for NHIs, workloads, automation, and AI agents.

Why it matters: It matters because IAM, PAM, and IGA teams now need to govern machine-driven privilege as a living access problem, not a one-time approval problem.

By the numbers:

• Non-human identities now outnumber human identities by as much as 144 to 1 in many organisations.
• The total number of NHIs continues to grow by more than 44% year over year.
• AI service credential leaks increased 81% in a single year.

Context

Traditional PAM was built around human administrators, where authentication, approval, and session boundaries are relatively clear. That model struggles when the dominant privileged actors are service accounts, workloads, pipelines, and AI agents that authenticate continuously and may exist only briefly.

The governance gap is not just volume. Machine identities rely on long-lived secrets, change behaviour at runtime, and can accumulate access faster than teams can recertify it. Once AI agents enter the picture, the question shifts from who approved access to what the identity can do while it is already in motion.

Key questions

Q: How should teams govern privileged non-human identities in agentic systems?

A: Treat the agent as a non-human identity with scoped authority, provenance controls, and explicit action authorization. Separate content reading from execution, require authenticated control channels for task changes, and remove any standing privilege the agent does not need. That governance model matters because prompt injection becomes far more dangerous once it can steer a privileged runtime.

Q: Why do long-lived secrets create more risk for infrastructure workloads?

A: Long-lived secrets are hard to inventory, rotate, and constrain across many workloads, so they tend to persist after they should be replaced. That increases the blast radius of compromise and makes governance harder in CI systems, cloud services, and Kubernetes. Short-lived, platform-issued tokens reduce standing exposure and better align access with the workload’s actual runtime context.

Q: What breaks when PAM only covers human administrators?

A: A human-only PAM model leaves service accounts, workloads, and AI-connected systems outside the same governance discipline. Those identities can still perform privileged actions, but they often bypass human approval, lifecycle review, and session oversight. That creates hidden privilege paths that are harder to audit and revoke.

Q: How should security teams compare zero trust and runtime identity control for NHIs?

A: Zero trust sets the policy posture, while runtime identity control is the enforcement pattern that applies it continuously to machine actions. For NHIs, the difference matters because the risk is not just initial access but whether each privileged action still meets policy after the session has started.

Technical breakdown

Why standing privilege fails for machine identities

Standing privilege assumes the identity is known, stable, and reviewable over a meaningful period. NHIs break that assumption because service accounts, APIs, and automation workflows can appear, act, and disappear on machine time rather than human time. Their credentials may be embedded, reused, or long-lived, which means the access path persists even when the original operational need has changed. In practice, that turns authentication into a durable attack surface instead of a temporary control point. Continuous operation also means the privilege boundary is crossed repeatedly, not just at login, so the control question becomes whether the identity still deserves access at the moment of each action.

Practical implication: move privileged access decisions from account creation to every meaningful runtime request.

How AI agents change the authorization model

AI agents are not just another workload because they can select actions at runtime, call tools, and continue without a human approving each step. That changes the identity problem from access possession to access use. Traditional policies assume the actor follows a predictable path, but an agent can vary tool choice and execution order as context changes. Even if the access was originally legitimate, the resulting activity may not stay within the operator’s intended scope. This is why continuous authorization matters more than initial authentication. The governing question is no longer simply whether the agent signed in, but whether each downstream action still matches policy, context, and purpose.

Practical implication: bind policy to each agent action, not to the session that started it.

What runtime identity control adds to PAM

Runtime identity control extends PAM from a gatekeeping function into an always-on authorization layer. Instead of granting broad standing access and reviewing it later, the control plane evaluates identity, target resource, and context throughout the life of the interaction. That is especially important for ephemeral workloads and AI-driven operations, where access can be both short-lived and high impact. The model does not replace least privilege. It changes where privilege is enforced, moving it closer to execution time. For practitioners, that means session auditability alone is not enough if the policy engine cannot constrain what happens after credentials are issued.

Practical implication: pair least privilege with runtime checks that can still intervene after credentials are issued.

Threat narrative

Attacker objective: The objective is to inherit persistent machine access and use it to reach sensitive resources, automate abuse, or widen the blast radius of privileged operations.

1. Entry occurs when machine identities authenticate with long-lived secrets, certificates, API keys, or service account passwords that remain valid well after their original purpose has changed.
2. Escalation follows when those identities retain standing privilege, allowing attackers or abused agents to perform actions beyond the original operational need without reapproval.
3. Impact emerges as privileged automation, cloud workloads, or AI agents touch sensitive systems at machine speed, widening the blast radius before human teams can intervene.

Breaches seen in the wild

• BeyondTrust breach 2024: A stolen BeyondTrust Remote Support API key let a China state-sponsored actor reset accounts and reach US Treasury workstations in 2024.
• reviewdog Action compromise 2025: A stolen maintainer token poisoned reviewdog/action-setup, leaking CI secrets including the tj-actions bot token used in the next attack.

Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Runtime identity control is now the correct PAM pattern for machine-driven enterprises. Static approval models do not match the way workloads, pipelines, and AI agents operate, because these identities authenticate continuously and can change behaviour faster than recertification cycles can react. The practical conclusion is that PAM has moved from session protection to execution-time governance.

Standing privilege is the central governance debt in NHI programmes. NHIs frequently accumulate access because their operational need is harder to observe than a human user’s job function. That makes long-lived credentials and broad entitlements the real source of drift, not merely weak hygiene. The practitioner takeaway is to treat access scope as a runtime variable, not a provisioning decision frozen at creation.

AI agents expose the assumption that authorised access stays aligned with intended action. That assumption was designed for systems that execute predefined steps and wait for human approval between major transitions. It fails when the actor can choose tools, sequence actions, and continue without a human gate, because the original approval no longer describes the behaviour. The implication is that governance must be rebuilt around action-level control rather than session-level trust.

Identity blast radius: when a machine identity can access multiple systems continuously, the compromise problem becomes one of propagation speed as much as privilege level. A credential that is valid across cloud services, APIs, and automation platforms can turn a single exposure into a cross-domain event. The practitioner conclusion is to map privilege breadth to potential propagation paths, not just to role names.

The market is converging on the idea that PAM, NHI governance, and AI agent control are part of the same control plane. The distinction between protecting a person, a service account, and an autonomous system is narrowing at the enforcement layer even when the identity types differ. That means teams should plan for shared policy, shared telemetry, and shared auditability across all three actor types. The conclusion is to stop treating machine privilege as a separate niche and start governing it as core identity infrastructure.

From our research library:

• NHIs outnumber human identities by 25x to 50x in modern enterprises, according to the Ultimate Guide to NHIs.
• Systems with least-privileged AI access had a 17% incident rate vs 76% for over-privileged systems. Organisations failing to scope AI access properly are 4.5x more likely to experience a security incident, according to the 2026 Infrastructure Identity Survey.
• Read next: Ultimate Guide to NHIs

What this signals

Runtime identity control: enterprises now need authorization that follows the action, not just the login, because workloads and AI agents can keep moving after the original access decision is obsolete. PAM teams should expect more demand for policy enforcement at execution time rather than at issuance time.

The governance shift is broader than secrets rotation. When machine identities can appear and disappear quickly, the durable control point becomes the policy layer that governs each privileged request, especially across CI/CD, cloud workloads, and AI-driven automation.

For practitioners

• Inventory privileged NHIs by runtime role Map service accounts, API keys, workloads, CI/CD pipelines, and automation platforms to the systems they can touch, then flag identities with standing privilege or unclear ownership.
• Replace standing privilege with task-scoped access Issue credentials for the shortest workable duration and bind them to a specific operational task, especially where workloads or automation need elevated access.
• Add continuous authorization to privileged sessions Evaluate identity, target resource, and context during execution, not just at authentication, so access can be reduced or denied while activity is still in flight.
• Separate human admin controls from machine governance Stop using administrator-centric PAM assumptions as the default model for NHIs and AI agents, because continuous authentication and autonomous action need different enforcement points.

Key takeaways

• Machine identities have become the main privileged actors in many environments, so PAM now has to govern software-driven access at runtime instead of only protecting human administrators.
• The core control gap is standing access that persists after the operational need changes, which gives compromised or over-scoped identities a wider blast radius.
• Teams should move toward task-scoped access, continuous authorization, and stronger ownership of NHIs to keep machine privilege aligned with actual use.

Key terms

• Runtime Identity Security: Runtime Identity Security is the practice of protecting identities while they are actively being used by software, services, devices, or agents. It focuses on controlling authentication, authorization, secrets, and session behavior at execution time, so compromised credentials, excessive privileges, or abnormal identity actions can be detected and contained quickly.
• Standing Privilege: Standing privilege is access that remains active even when no immediate task requires it. For NHI programmes, it is a common failure mode because long-lived credentials and persistent roles create unnecessary exposure. Reducing standing privilege usually means tighter expiry, on-demand access, and clearer review of who or what still needs access.
• Non-Human Identity (NHI): A digital identity assigned to a non-human entity such as a software application, service account, API key, bot, machine, or AI agent that enables it to authenticate and interact with systems without direct human involvement. NHIs now outnumber human identities in most enterprises by 25 to 50 times.
• Continuous authorization: Continuous authorization is the practice of rechecking access as a session unfolds instead of trusting a single login decision. It matters for AI workflows because the request, context, retrieved data, and downstream action can all change between prompt and execution, making static approval too blunt.

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 responsible for identity security strategy or NHI governance in your organisation, it is worth exploring.