Continuous authorization for AI agents and NHIs is the new baseline

At a glance

What this is: Delinea argues that identity security is moving to continuous, context-aware authorization that follows identities through the session, not just the login.

Why it matters: That matters because IAM, PAM, and NHI programmes have to govern access after issuance as well as before it, especially as machine identities and AI agents expand the attack surface.

By the numbers:

• Delinea customers see an 88% reduction in the overhead associated with provisioning and deprovisioning users.
• The vendor reports a 90% reduction in alert noise through AI-driven prioritization.

👉 Read Delinea's analysis of continuous authorization for identity security

Context

Continuous authorization is the idea that access decisions should keep evaluating what an identity is doing after the session starts, rather than stopping at initial authentication or approval. That matters for AI agent identity, machine identity, and privileged human access because the risk is often created by what happens during execution, not at sign-in.

The governance gap is that many IAM and PAM programmes still treat access as a one-time event. In environments where third parties, workloads, and AI agents move quickly across systems, static approval flows and long-lived privileges do not match operational reality. The question is no longer whether access was granted correctly, but whether it stayed justified at every step.

Key questions

Q: How should security teams implement just-in-time access for privileged identities?

A: Start by limiting just-in-time issuance to tasks with clear scope, short duration, and measurable risk. Pair the request with a business justification, revoke access automatically at session end, and verify that logs show both issuance and removal. The goal is to make standing access the exception, not the operating model.

Q: Why do machine identities make continuous authorization harder to manage?

A: Machine identities often operate at higher speed and volume than human users, which means privilege can be created, used, and reused faster than manual governance can observe it. Continuous authorization is harder because the control must track behaviour in session, not just the identity record. That requires tighter policy telemetry and stronger audit trails.

Q: What breaks when privileged access stays active after the task is done?

A: The main failure is that the access window outlives the business need, so any compromise, misuse, or mistake can continue without a new approval step. That increases exposure for human admins, service accounts, and AI-driven workflows alike. Mature programmes close the gap by revoking access as soon as the task is complete.

Q: How do teams know whether continuous authorization is actually working?

A: Look for evidence that sensitive actions are being checked in real time, that false positives are being reduced without weakening enforcement, and that every decision can be explained after the fact. If teams still rely on manual review after the session, the control is not continuous in practice. Auditability is part of effectiveness, not separate from it.

Technical breakdown

Zero Standing Privilege and just-in-time access

Zero Standing Privilege means identities do not retain permanent access between tasks. Instead, permissions are issued just in time, scoped to a specific action, and revoked when the session ends. For privileged humans this reduces persistent exposure. For NHIs and AI agents, the same principle limits how long tokens, secrets, or elevated rights remain usable if an identity is compromised. The operational value is not only reduced standing privilege, but also a narrower window for misuse when execution begins.

Practical implication: treat long-lived privileged access as an exception and require explicit task-scoped issuance for high-risk identities.

Continuous authorization at the execution layer

Continuous authorization evaluates each command, query, or configuration change against current policy while the session is active. This is different from perimeter access control because the decision is made at the moment of action, using context such as identity, task, and expected behaviour. That design matters for NHI and agentic workflows because an identity can start within policy and later drift into a higher-risk action path. The control point moves from login to execution, where the actual damage is usually created.

Practical implication: enforce policy where privileged actions occur, not only at authentication or initial approval.

Explainable AI-driven authorization

AI-driven authorization can help prioritise alerts and surface anomalies, but only if the reasoning is explainable enough for audit and response teams to trust it. In identity governance, black-box scoring is not enough because teams need to know what evidence triggered a decision, what changed, and what was allowed or blocked. That is especially relevant for machine identities and AI agents, where volume and speed make manual review impossible. Explainability turns enforcement from a hope into something a board, auditor, or incident team can verify.

Practical implication: require decision evidence and audit trails for any automated authorization that can change privileged access.

Threat narrative

Attacker objective: The attacker aims to turn legitimate identity access into sustained operational control without triggering a timely privilege boundary.

1. Entry occurs when an identity receives access that is valid at login but remains broadly usable for the rest of the session.
2. Escalation happens when the identity keeps enough privilege to issue sensitive commands, call protected systems, or retain secrets after the original task has changed.
3. Impact is created when over-provisioned access or slow enforcement allows configuration change, data exposure, or lateral movement before review catches up.

Breaches seen in the wild

• Moltbook AI agent keys breach — Moltbook breach exposed 1.5M AI agent keys.
• Cisco DevHub NHI breach — IntelBroker exploited exposed Cisco credentials, API tokens and keys in DevHub.

Read our 52 NHI Breaches Analysis report for a comprehensive view of breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Continuous authorization is becoming the defining control plane for modern identity governance. Static approval workflows assume the risk decision happens once, but privileged work now unfolds across chained systems, secrets, and sessions. That makes the authorization moment continuous, not episodic. Practitioners should treat session-level control as a core IAM and PAM requirement, not an add-on.

Zero Standing Privilege is no longer just a privileged access pattern, it is a response to identity speed. The article’s core claim is that access should exist only when the task exists, because persistent entitlements create avoidable exposure. That position aligns with OWASP-NHI and Zero Trust thinking for machine identities, third parties, and privileged operators. Practitioners should re-evaluate any programme that still relies on durable access as the default.

Explainable automation matters because identity control now has to be defensible as well as effective. AI-assisted authorization can reduce noise, but it also shifts more trust into policy models and decision logic. When those decisions affect privileged humans, workloads, or AI agents, auditors will ask why something was allowed, not just whether it was blocked. Practitioners should insist on evidence-backed authorization decisions that can survive review.

Continuous controls expose the weakness of approval-centric governance for high-speed identities. Board-level comfort often comes from seeing requests approved, but approval is not containment. The article shows that business acceleration and security are becoming the same problem space, because identity policy now has to keep pace with execution. Practitioners should judge maturity by in-session enforcement, not by how quickly tickets move.

From our research:

• Two-thirds of enterprises have endured a successful cyberattack resulting from compromised non-human identities, with a quarter encountering multiple attacks, according to The 2024 ESG Report: Managing Non-Human Identities.
• Enterprises that have experienced a compromised NHI averaged 2.7 separate incidents in the past 12 months.
• For a broader view of the failure modes behind those incidents, see 52 NHI Breaches Analysis.

What this signals

Identity programmes are shifting from request governance to runtime governance. The practical question is no longer whether access was approved, but whether the control could still intervene once the session started. Teams that rely on ticket closure or manual certification will keep missing the moment of misuse. For a wider reference point on NHI risk patterns, compare this with the Ultimate Guide to NHIs , Key Challenges and Risks.

Continuous authorization will force IAM, PAM, and NHI teams into the same operating model. The more identities behave like workloads or agents, the less useful it becomes to separate access issuance from access enforcement. That convergence will favour programmes that can correlate entitlement, session, and action in one control plane. The OWASP OWASP Non-Human Identity Top 10 remains a useful lens for the underlying exposure patterns.

AI-assisted policy enforcement will raise the bar for explainability, not just automation. The interesting question for practitioners is whether their controls can prove why a command was allowed, not whether they can generate a score. With 72% of organisations having experienced or suspecting a breach of NHIs, according to the 2024 ESG Report: Managing Non-Human Identities, runtime evidence becomes a governance requirement.

For practitioners

• Map privileged access to task boundaries Identify where human administrators, service accounts, and AI agents still keep access after the task is finished. Replace durable entitlements with task-scoped issuance and require explicit revocation at session end.
• Enforce policy at the execution layer Place controls where commands and configuration changes happen, not only at sign-in. Use real-time checks for high-risk actions so a valid session cannot silently drift into unauthorized activity.
• Require evidence for automated decisions Demand that any AI-assisted authorization decision records the context, policy trigger, and outcome in a way that can be reviewed by operations, audit, and incident response teams.
• Reduce long-lived privileged exposure Audit which secrets, tokens, and admin pathways remain usable across multiple sessions. Move those paths toward just-in-time access and short-lived credentials wherever operationally possible.

Key takeaways

• The article’s core argument is that identity security now has to govern the session, not just the login.
• The strongest evidence is operational, not theoretical: continuous controls cut privilege exposure, alert noise, and provisioning overhead.
• Practitioners should measure maturity by how fast they can enforce, explain, and revoke access during execution.

Key terms

• Zero Standing Privilege: A governance model where identities do not keep permanent elevated access. Privileges are issued only when needed, for a specific task, and removed as soon as the work is complete. The control reduces the exposure window for both human and non-human identities.
• Continuous Authorization: A control approach that keeps evaluating access while a session is active. It checks whether an action still matches policy at the moment of execution, rather than assuming the original grant remains valid for the entire session.
• Execution Layer: The point in the workflow where commands, queries, or configuration changes are actually carried out. For identity governance, this is where the most useful control can be applied because it sees the action itself, not just the request for access.
• Explainable Authorization: An authorization model that can show why a decision was allowed, blocked, or flagged. In identity operations, explainability matters because audit, incident response, and security leadership all need evidence they can verify instead of opaque scoring.

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.

This post draws on content published by Delinea: Continuous. Contextual. Controlled. The new standard for identity security. Read the original.