Agentic traffic is forcing a new identity security model

At a glance

What this is: This is a conference reflection arguing that agentic traffic and AI adoption are forcing security teams to move from perimeter thinking to identity proof and action-level authorisation.

Why it matters: It matters because IAM, PAM and NHI programmes now have to govern legitimate machine and agent behaviour, not just authenticate users and workloads.

By the numbers:

• agentic traffic is still less than 10% of total volume

Context

AI adoption is no longer just about adding automation. The security problem is that organisations are now trying to govern legitimate action across humans, applications, agents and backend workloads at the same time, while the old origin-based trust model becomes less useful.

That shift creates an identity governance gap rather than only a detection gap. When the same environment has to distinguish between human intent, workload intent and agentic behaviour, access decisions need to move closer to the action itself, especially where least privilege for AI agents is not yet established.

Key questions

Q: How should security teams govern AI agents that can access enterprise systems?

A: Security teams should govern AI agents as non-human identities with explicit ownership, scoped privileges, and continuous monitoring. The control set should include inventory, task-bound credentials, audit trails, and revocation paths. If an agent can call tools or touch production systems, it belongs in the same governance model as service accounts and other machine identities.

Q: Why do least-privilege models need to change for agentic AI?

A: Least privilege still matters, but agentic AI changes the timing and granularity of enforcement. A static entitlement set assumes the actor’s purpose is stable enough to review later. Agents can request, combine and use access inside a single task, so privilege has to be bounded by purpose, time and action rather than by broad role assignment.

Q: How should security teams verify identity when AI-generated profiles look authentic?

A: Security teams should stop relying on visual realism or message quality as proof of identity. Stronger verification should combine liveness checks, device and account signals, behavioural risk scoring, and step-up review at moments where trust is being extended. The goal is to verify the claimant, not the quality of the synthetic content.

Q: How do organisations decide whether an AI agent should be allowed to act autonomously?

A: Autonomy should be granted only when the business process can tolerate runtime action, the owner is clear, the policy engine can make immediate decisions, and the activity is fully attributable. If any of those are missing, the safer model is constrained delegation with challenge or denial for sensitive steps.

Technical breakdown

Why origin-based bot detection breaks down

The article’s central technical point is that identity checks built around source, reputation or simple bot classification are no longer enough. Agentic traffic can look legitimate, operate at scale and still be misaligned with policy. That means the control problem is no longer only “who is this?”, but “is this action authorised for this identity type in this context?” In practice, this pushes security teams toward stronger intent verification, tighter runtime authorisation and proofs that are bound to the specific request rather than the channel alone. See the AI Agent Authorisation Guide.

Practical implication: move from source-centric detection to action-centric authorisation for agents, workloads and users.

How cryptographic proof anchors identity in mixed environments

Cryptographic proof is presented here as the foundation for a mixed ecosystem where a person, app, agent or workload may all be acting through digital identity. The key issue is not just authentication at login or startup, but proving the communicating entity and its intent at the point of use. That aligns with zero trust thinking: trust is continuously evaluated, not assumed after a single check. For agentic systems, this is especially important because runtime behaviour can change without a human in the loop. See the Identity Governance for AI Agents Guide.

Practical implication: bind authentication, authorisation and session validity to each transaction where agent behaviour can change mid-flow.

Why least privilege must become task-scoped for agents

The post treats zero trust, least privilege and continuous authentication as the operating model for agentic environments. That works only if privilege is treated as temporary and bounded to a specific task, not as a stable entitlement handed to an entity and assumed to remain safe. The architectural problem is that agentic behaviour can create more requests, more tool use and more opportunities for overreach than human users usually do. Governance therefore has to constrain what an agent can do in the moment, not just what it was allowed to do at registration.

Practical implication: define task-scoped permissions for agents and revoke access when the task ends.

NHI Mgmt Group analysis

Agentic traffic creates an identity governance problem, not just a detection problem. The article’s strongest implication is that security teams can no longer treat bots, workloads and agents as a single behavioural class. A system that can take actions autonomously or semi-autonomously needs governance that understands who or what is acting, under what authority, and for which task. That is a different control question from traditional perimeter filtering, and it is why programme owners should separate identity assurance from traffic classification.

Action-level authorisation is becoming the real control plane for AI-driven environments. The old question was whether traffic was malicious or not. The new question is whether a legitimate identity is allowed to perform a legitimate action in this exact moment. That pushes governance closer to runtime, with policy decisions tied to intent, scope and session state. For practitioners, this is where identity, PAM and workload controls start converging into one operating model.

Identity proof without intent proof is incomplete in agentic ecosystems. The article is right to elevate cryptographic proof, but proof of entity alone does not solve the governance problem if the action itself is unconstrained. Enterprises need to think in terms of identity plus declared purpose plus bounded authority. That framing matters most when agents, apps and humans all use the same digital fabric, because the programme must prove not only who is there but what they are authorised to do.

Least privilege has to be reinterpreted as transient privilege for non-human actors. The article points to zero trust, least privilege and continuous authentication as the right direction, but the deeper issue is that static entitlement models do not map cleanly to agentic behaviour. When privilege can be expanded by runtime context, governance has to assume the action can change faster than review cycles can observe it. The practitioner takeaway is that entitlement design must follow execution tempo, not human admin cadence.

From our research library:

• Organisations that describe themselves as confident in their AI deployment actually experience a 72% security incident rate, compared to 33% for those who remain cautious, according to the 2026 Infrastructure Identity Survey.
• Only 13% of organisations feel extremely prepared for the reality of agentic AI despite the majority racing toward autonomous adoption, according to the 2026 Infrastructure Identity Survey.
• Read next: Identity Governance for AI Agents Guide

What this signals

Identity governance for agents is moving from lifecycle administration to runtime control. Programmes built around periodic review will struggle if agents can obtain and use access faster than a human can certify it. That is why the next maturity step is to govern issuance, purpose and expiry together rather than treat access review as the primary safeguard.

Action-level policy will matter more than source-based trust. As environments absorb more AI-driven behaviour, the decisive question becomes whether an identity may perform a specific action right now. The security model shifts from origin trust to transaction trust, which is why architectures that cannot express task-scoped authority will fall behind.

Confidence in AI adoption does not equal operational readiness. Organisations that describe themselves as confident in their AI deployment actually experience a 72% security incident rate, compared to 33% for those who remain cautious, according to the 2026 Infrastructure Identity Survey. The signal for practitioners is clear: adoption speed needs to be matched by identity governance depth, or incident rates will rise with scale.

For practitioners

• Separate agent governance from generic bot monitoring Treat AI agents, automation and ordinary bots as different control classes. Document which identities can initiate actions, which can only respond, and which require task-scoped limits.
• Bind authorisation to each action request Move policy decisions closer to the transaction so that an agent, workload or user is checked against purpose, scope and session context before the action runs.
• Inventory where cryptographic proof is still missing Map the systems that rely on channel trust, static tokens or weak origin checks and identify where stronger proof of entity and intent is required.
• Redesign review cycles for non-human behaviour Assume agents may create, combine and retire privileges faster than periodic reviews can catch. Shift governance toward issuance-time controls and short-lived authority.

Key takeaways

• Agentic traffic is forcing security teams to govern behaviour at the action level, not just authenticate the source of a request.
• The article says less than 10% of total traffic can already reshape security priorities when it is agentic and influential.
• Identity proof, least privilege and continuous authentication only hold up if they are redesigned for non-human behaviour and runtime decisions.

Related NHI Mgmt Group resources: Identity Governance for AI Agents Guide · AI Agent Authorisation Guide · Agentic AI Identity Maturity Model · Agentic AI Identity Guide

Key terms

• Agentic Traffic: Traffic generated by software that can act on behalf of a user or process with some degree of independent decision-making. In fraud prevention, it includes both legitimate assistants and malicious automation, so the control question becomes intent and behaviour, not automation alone.
• Resource-Level Authorisation: A control pattern where each application or service decides whether a subject should be allowed to act. The decision is based on identity, context, and policy rather than on network location. This is the practical mechanism that makes zero trust enforceable in real environments.
• Cryptographic Proof: A method of proving possession or authenticity using mathematically verifiable evidence rather than remembered knowledge or easily copied personal data. In identity workflows, cryptographic proof helps confirm that a user or device is linked to the claimed identity with a much higher level of assurance.
• Task-Scoped Least Privilege: A permission model that limits an agent or service identity to only the systems and data required for a specific task. It reduces exposure by preventing a non-human system from carrying unnecessary access into every workflow it executes.

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.