TL;DR: Privilege is shifting from vaulting admin accounts to governing actions taken by service accounts, automation pipelines and AI agents, according to Delinea’s summary of two 2026 analyst reports. Runtime authorisation, not credential storage alone, is now the control point that determines whether privilege is actually constrained.
At a glance
What this is: This is a Delinea analysis arguing that PAM is shifting from protecting admin accounts to governing actions at runtime across machines and AI agents.
Why it matters: It matters because PAM, IGA and secrets teams now have to control what identities are allowed to do in real time, not just who holds credentials.
Context
PAM is no longer only about locking down administrator accounts. The governance problem has shifted to controlling actions taken by non-human identities, including service accounts, automation pipelines and AI agents, because those identities can act continuously and outside human work patterns.
Traditional vaulting and session recording still help, but they do not decide whether a specific action should be allowed at the moment it is taken. That makes runtime authorisation the control point that matters most for privileged access programmes built around modern machine and agent behaviour.
Key questions
Q: What breaks when PAM still centers on accounts instead of actions?
A: Account-centric PAM misses the decision point that now matters most. When service accounts, pipelines and AI agents are the ones acting, a vault can protect the secret but still fail to stop an unauthorised action. The control must move to the moment of execution, where the policy can permit or deny the action itself.
Q: Why do standing privileges make machine identities harder to secure?
A: Standing privileges extend the period in which a compromised secret can be abused, which enlarges the blast radius of a single exposure. They also make it harder to prove who had access, when it was used, and whether it should still exist, especially in cloud environments with high identity sprawl.
Q: How can organisations tell whether runtime authorization is actually working?
A: Look for three signs: decisions happen fast enough to stay inline, policies use live context instead of stale claims, and every allow or deny produces an auditable record. If teams cannot explain a specific decision after the fact, or if applications bypass the control because it is too slow, the runtime layer is not functioning as intended.
Q: When should teams prioritise just-in-time access over session recording?
A: Prioritise just-in-time access when the risk comes from standing privilege, fast-moving cloud operations, or machine-speed workflows that cannot wait for later review. Recording can supplement governance, but it should not be the first line of defence when the goal is to prevent privileged misuse before the session begins.
Technical breakdown
Why account-centric PAM misses runtime decisions
Classic PAM assumes privilege can be identified by account, credential and session. That model works when humans log in episodically and their actions are easy to tie back to a named administrator. It breaks when service accounts, automation pipelines and AI agents are the active privilege holders, because their authority is expressed through actions, not just through login events. A vault can store a secret and session recording can replay behaviour, but neither control decides whether the action itself was authorised. The real control boundary moves from credential custody to per-action policy enforcement.
Practical implication: map privileged workflows to the action that is being authorised, not only to the account that initiates it.
Just-in-time access and zero standing privilege for machines
Just-in-time access and zero standing privilege become more important when identities operate at machine speed. Instead of issuing a reusable credential that persists between tasks, the programme grants access only when a task requires it and removes it when the task ends. That reduces the time a secret exists in a place an attacker can steal it, and it limits the duration of privilege if the workload or agent is compromised. In this model, privilege is expressed as an ephemeral allowance tied to a specific action window, not as a durable entitlement attached to the identity.
Practical implication: replace persistent machine credentials with task-scoped access windows wherever the workload can tolerate it.
Runtime authorisation is now the control plane for privilege
Runtime authorisation means the access decision happens when the identity attempts the action, using context such as target, timing and scope. That is different from pre-provisioned access, where the identity already holds broad permission and the control only observes after the fact. For AI agents and continuously running workloads, the difference matters because privilege can expand and contract across many actions in a short period. PAM therefore becomes a policy engine for what the identity may do right now, rather than a repository for secrets alone.
Practical implication: enforce policy decisions at action time so that governance keeps pace with machine and agent execution.
Threat narrative
Attacker objective: The attacker objective is to turn durable machine privilege into repeated unauthorised actions before the access model can intervene.
- Entry occurs when a service account, automation pipeline or AI agent receives a reusable privileged credential that outlives the task it was meant for.
- Escalation follows when that credential can be used for many actions without a fresh policy decision, allowing broad privileged behaviour at machine speed.
- Impact is the ability to perform unauthorised administrative actions, because the control plane observed the identity but never constrained the action itself.
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.
- Uber breach 2022: A contractor's stolen password and MFA fatigue gave a Lapsus$-linked attacker Uber's internal tools; Uber rotated keys to many services.
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
Privilege has moved from accounts to actions, and PAM has to follow it. The old unit of control was the admin account, but the real privilege boundary now sits inside the action itself. That shift matters because service accounts, automation pipelines and AI agents can carry privilege without looking like classic administrator sessions. Practitioners should treat runtime authorisation as the core control model, not an optional extension of vaulting.
Vaults and session recording remain necessary, but they are no longer sufficient. A vault can protect a secret and a session recorder can show what happened after access is granted. Neither control answers whether the action should have been permitted at that moment. That is why the modern PAM discussion increasingly overlaps with identity governance and workload identity, where the question is permission to act rather than possession of a password.
Just-in-time privilege is becoming the default governance pattern for non-human execution. When identities operate continuously, standing access creates unnecessary exposure and weakens accountability. Ephemeral access tied to the task reduces the opportunity for credential reuse and broadens the gap between identity persistence and privilege persistence in a useful way. Teams should expect runtime controls to define the next phase of PAM maturity.
Runtime control is where PAM, secrets management and identity governance converge. The market is moving toward platforms that can discover identities, evaluate risk and decide access in the moment of action. That convergence does not eliminate existing controls, but it does change which layer becomes decisive during review and enforcement. The practical conclusion is that privilege programmes now succeed or fail on the quality of their real-time decision path.
From our research library:
- 97% of NHIs carry excessive privileges, increasing unauthorised access and broadening the attack surface, according to the Ultimate Guide to NHIs.
- Read next: Just-in-Time Access and Zero Standing Privilege Guide
What this signals
Privilege is becoming a runtime governance problem. Programmes that still treat privileged access as a credential storage issue will miss the control point that now determines risk. The decisive question is whether an identity may perform a specific action at the moment it attempts it, which is why action-level policy has to sit alongside vaulting and audit.
Task-scoped access will matter more than durable entitlements. For machine and agent execution, standing privilege creates unnecessary exposure between jobs, while ephemeral access narrows the window in which a credential can be reused. That makes just-in-time control a programme design choice, not a niche optimisation.
Access reviews assume privileges persist long enough to be certified. Service accounts and AI agents can execute and release access faster than many review cycles operate, so governance has to shift earlier in the lifecycle, at issuance and action approval.
For practitioners
- Map privilege to actions, not just accounts Inventory the high-risk actions performed by service accounts, pipelines and AI agents, then identify where today’s controls only govern the credential rather than the action decision.
- Reduce standing privilege for non-human identities Replace durable machine credentials with task-scoped access where the workload can support it, so access exists only for the window needed to complete the action.
- Separate observation from enforcement Keep session recording and audit trails, but add a runtime authorisation layer that can approve or deny the action before it executes.
- Unify discovery across human and machine privilege Build a single inventory that shows which identities, human and non-human, can perform which privileged actions, and review it from the same governance process.
Key takeaways
- Privilege governance is moving from account custody to action control, especially where machines and AI agents are the active identities.
- Traditional PAM controls still matter, but they cannot determine whether a specific privileged action should be allowed in the moment.
- The practical shift is toward runtime authorisation, just-in-time access and lower standing privilege across non-human execution paths.
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 MITRE ATT&CK address the attack and risk surface, while NIST SP 800-53 Rev 5 and NIST CSF 2.0 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-05 — Overprivileged NHI | The article centers on privileged access for service accounts, pipelines and AI agents. |
| NHI-07 — Long-Lived Secrets | The post argues that persistent credentials no longer fit continuous machine and agent execution. | |
| Recommendation — Reduce overprivilege by moving machine access to task-scoped permissions and eliminating durable entitlements. Replace long-lived secrets with ephemeral access paths that expire with the action window. | ||
| NIST SP 800-53 Rev 5 | IA-5 — Authenticator Management | Authenticator lifecycle is central when PAM shifts from vaulting accounts to governing runtime access. |
| Recommendation — Apply authenticator management to rotate, scope and revoke privileged credentials tied to machine actions. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | Runtime privilege control maps directly to permissions and authorisations management. |
| Recommendation — Control entitlements at execution time and verify that privileged actions are explicitly authorised. | ||
| MITRE ATT&CK | TA0006;TA0008 — Credential Access; Lateral Movement | The threat pattern is durable credential abuse leading to broader privileged activity. |
| Recommendation — Map machine credential exposure and repeated privileged action paths to credential access and lateral movement techniques. | ||
Key terms
- Runtime Authorisation: Runtime authorisation is the practice of deciding access while a task is in progress, rather than only at provisioning time. It matters for NHIs because credentials and entitlements can change risk mid-session, especially when automation or AI agents interact with sensitive systems.
- Just-in-Time Access Request: Just-in-Time Access Request is a pattern that grants access only when it is needed and only for the duration required. It reduces standing privilege by making access temporary, policy driven, and task scoped. This approach is especially useful for contractors, sensitive systems, and short-lived operational work.
- Zero Standing Privilege: A control model in which an identity does not keep persistent access unless it is actively needed. For NHIs, this means credentials and permissions are issued for a narrow task and then removed. It reduces the time window and reuse value of stolen access.
- Action-Centric PAM: A PAM operating model that governs the actions an identity may take, not just the account it uses. This matters when privilege is exercised by service accounts, automation pipelines or AI agents that do not fit human session patterns.
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.
Published by the NHIMG editorial team on July 9, 2026.
Updated on October 11, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org