Pathlock CEO on identity security in the AI era

At a glance

What this is: This is a short Pathlock interview about how agentic AI is reshaping identity security, governance, and compliance for enterprises and channel partners.

Why it matters: It matters because IAM, IGA, PAM, and NHI programmes now have to govern machine and autonomous access patterns that do not fit human-era review, provisioning, and accountability assumptions.

👉 Read Pathlock's interview on identity security in the AI era

Context

Agentic AI changes identity security because it can make runtime decisions about tools, data, and execution paths instead of waiting for a human request. That creates a governance problem for programmes built around static entitlements, predictable approval flows, and access review cycles. For identity teams, the question is no longer only who has access, but what kind of actor is making the access decision.

Pathlock frames the issue through identity, governance, and compliance, which is the right lens for enterprise security teams. Once AI systems begin acting as decision-making actors inside business workflows, IAM, IGA, and PAM controls need to be evaluated against the behaviour of the actor, not just the system record. That makes agentic AI an identity governance problem as much as an AI risk problem.

Key questions

Q: What breaks when agentic AI is governed like a normal application?

A: Normal application governance assumes the system follows a fixed path once access is granted. Agentic AI can decide which tool to use, when to act, and how to continue, so static approval models miss the real control point. Teams need governance around runtime decision authority, not only application entitlements.

Q: Why do agentic AI systems complicate IAM and IGA programmes?

A: They complicate IAM and IGA because the actor can exercise access dynamically rather than through a stable, human-paced workflow. That means recertification, SoD, and exception handling may all occur after the action has already happened. The control issue is timing, not just scope.

Q: How should organisations govern privileged access for AI-driven workflows?

A: They should treat AI-driven workflows as bounded execution paths with explicit approval, logging, and scope limits. Privilege should be tied to a task, a tool set, and a responsible owner, with clear rules for when the system can delegate or escalate. PAM and policy enforcement need to work together.

Q: Who is accountable when an autonomous AI system acts outside intent?

A: Accountability should remain with the business owner of the workflow, the team that approved the AI’s privileges, and the control owners responsible for monitoring. If those roles are unclear, the governance model is already too weak. Autonomous behaviour does not remove accountability, it exposes where it was never assigned.

Technical breakdown

Agentic AI as an identity actor

Agentic AI becomes an identity problem when the system can choose actions, select tools, and decide when to execute without a human gate in the loop. At that point, the actor is no longer just consuming permissions, it is operationalising them. Traditional IAM models assume the identity is authenticated and then constrained by pre-set policy. Agentic behaviour breaks that neat separation because the actor can alter its path through a workflow at runtime, which changes how privilege, accountability, and auditability must be interpreted.

Practical implication: treat autonomous AI behaviour as an identity subject that needs explicit governance boundaries, not just application access.

Why governance and compliance controls strain under agentic workflows

Governance controls such as access certification, segregation of duties, and approval workflows were designed for identities whose privileges remain stable long enough to be reviewed. Agentic systems compress those assumptions because action, delegation, and execution can happen inside a single task cycle. That means evidence collection may lag behaviour, and compliance artifacts may only show that access existed, not how it was used. The core issue is not visibility alone, but whether the control model can observe and govern runtime delegation.

Practical implication: align certification, logging, and approval workflows to the actual runtime behaviour of AI-enabled processes.

What changes for PAM and machine identity programmes

Agentic AI increases the pressure on PAM and NHI programmes because privileged access is no longer limited to humans or fixed service accounts. The identity can be ephemeral, delegated, or orchestrated across tools, which makes entitlement scope and revocation timing more complex. In practice, that pushes teams toward tighter task scoping, stronger policy boundaries, and better mapping of who or what is allowed to act on behalf of the business. The control objective shifts from simple access grant to bounded execution.

Practical implication: review privileged access paths for AI-driven processes and define clear execution boundaries before deployment.

NHI Mgmt Group analysis

Agentic AI turns identity from a static entitlement problem into a runtime governance problem. The key shift is that access is no longer just granted and reviewed, it is exercised by an actor that can decide how to proceed inside the workflow. That changes the control question from entitlement ownership to decision authority. Practitioners should reframe governance around runtime behaviour, not only access records.

Identity review cycles were built for access that persists long enough to be certified. Agentic systems can obtain, use, and release privilege inside a task window that may not line up with review cadence. That makes conventional recertification weaker as a control signal because the evidence trail may outlive the decision event. The implication is that IAM teams must rethink what constitutes reviewable access in the first place.

Executive oversight of agentic AI belongs in the same governance stack as PAM and IGA. The article’s main value is that it connects AI-era identity issues to established governance disciplines rather than treating them as a separate AI-only concern. That is the right model for programme design because accountability, approval authority, and privilege boundaries all converge when an autonomous system can act on behalf of the enterprise. Practitioners should align AI governance with identity governance, not run them as separate tracks.

Runtime delegation is the named concept identity teams now need to manage. Once an AI system can chain decisions across tools and workflows, the problem is not simply access abuse, but delegated action that expands beyond the original approval intent. That is a different failure mode from ordinary machine identity sprawl. Practitioners should map where delegated execution begins, where it ends, and who remains accountable when it does not behave as planned.

From our research:

• 97% of NHIs carry excessive privileges, increasing unauthorised access and broadening the attack surface, according to Ultimate Guide to NHIs.
• 71% of NHIs are not rotated within recommended time frames, increasing the risk of compromise over time.
• 52 NHI Breaches Analysis shows how credential exposure and standing privilege repeatedly turn into real incidents.

What this signals

Runtime delegation is becoming the new control boundary: as agentic systems move from experimentation into production workflows, security teams will need to govern execution paths rather than only accounts. That shifts the programme focus from provisioning completeness to decision containment, especially where privileged access can be exercised faster than review cycles can observe it.

Pathlock’s interview reinforces a broader industry pattern: AI governance and identity governance are converging, and teams that keep them separate will miss the operational failure points. Organisations that already struggle with NHI privilege sprawl should expect the same control debt to appear in autonomous workflows, only with less predictable timing and weaker human traceability.

With 97% of NHIs carrying excessive privileges, according to our NHI reference research, the issue is not whether privileged access can be reduced later. The real challenge is whether current governance stacks can distinguish between a permission, a delegation, and an autonomous action before the actor has already executed.

For practitioners

• Map agentic decision points Identify every workflow where an AI system can choose actions, tools, or timing without a human approval gate. Classify those points as identity control boundaries so IAM and governance teams can assign ownership and log requirements.
• Separate reviewable access from runtime use Update access review processes so they distinguish between permission to act and actual execution by the actor. If the system can complete a task before the next certification cycle, add runtime controls and event-based evidence collection.
• Bound privileged AI workflows Define explicit ceilings for delegated privilege, approved tool sets, and task scope before deploying agentic systems into production. Tie those boundaries to PAM, IGA, and policy enforcement rather than informal operating assumptions.

Key takeaways

• Agentic AI shifts identity security from static entitlement control to runtime governance of decisions, tools, and execution paths.
• Conventional review and certification cycles are a poor fit when AI-enabled actors can obtain and use privilege within a single task window.
• IAM, IGA, and PAM teams should define bounded execution, accountable ownership, and event-level evidence before agentic systems reach production.

Key terms

• Agentic AI: Software that can choose actions, tools, and timing at runtime rather than following only pre-scripted automation. In identity programmes, agentic AI behaves like a decision-making actor that can consume privileges dynamically, which makes governance, auditability, and accountability materially harder than for ordinary workflows.
• Runtime delegation: The transfer or use of authority during execution, rather than at provisioning time. For autonomous systems, runtime delegation means the actor can extend, chain, or narrow its own action path in ways that may not match the original approval intent, creating a governance gap between policy and behaviour.
• Identity governance: The discipline of defining, approving, reviewing, and revoking access in a way that matches business risk and accountability. For AI-enabled environments, identity governance has to cover not just who holds access, but what kind of actor is using it and whether its behaviour can change mid-session.
• Privileged access: Access that can materially affect systems, data, or business operations if misused. In agentic contexts, privileged access is not limited to human administrators, because autonomous systems can inherit, select, and exercise privileged capabilities across tools, which makes containment and ownership essential.

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 Pathlock: Pathlock CEO Talks Identity in the AI Era. Read the original.