TL;DR: As enterprises add co-pilots, autonomous assistants and bots into core workflows, the old assumption that one user equals one accountable identity no longer holds, according to Gathid. Governance now has to address delegated and autonomous actions as part of the identity surface, not as after-the-fact exceptions.
At a glance
What this is: Gathid argues that human-machine collaboration is breaking the old one-user, one-accountability identity model and forcing organisations to treat delegated and autonomous actions as part of the identity surface.
Why it matters: This matters because IAM, IGA and PAM programmes now have to assign ownership, logging and revocation across human, delegated and autonomous actors, not just human users.
Context
Enterprise identity programmes were built on a simple premise: a user is a person, and the person is the accountable actor. That premise fails when co-pilots, assistants and bots can act inside business workflows with their own access, their own permissions and their own operational boundaries. The governance problem is no longer just authentication. It is accountability across a distributed identity surface.
The article frames this as a human-machine identity problem, not a pure AI tooling problem. In practice, the same workflow can now involve a human, delegated automation and autonomous action, which means identity ownership, access review and revocation logic all need to be able to distinguish who acted, what was delegated and what was substituted.
Key questions
Q: What breaks when human and machine identities share the same workflow?
A: The one-user, one-accountability model breaks down. A workflow can now involve a person, delegated automation and autonomous action, so ownership, audit evidence and revocation must follow the whole identity chain instead of a single login event.
Q: Why do delegated bots and assistants create accountability risk?
A: Because they act with permission but not always with the same intent or context as the human who enabled them. That makes it harder to prove who owned the action when the outcome becomes disputed, harmful or subject to review.
Q: How should security teams govern machine identities differently from human users?
A: Security teams should govern machine identities with lifecycle, context, and runtime controls, not human approval workflows. That means separate ownership, purpose-based entitlement, expiry, and revocation for service accounts, bots, and agents. Human IAM can inform the model, but machine access needs faster review, tighter scoping, and automated enforcement.
Q: What should organisations do when an employee leaves but the automation remains?
A: They should retire the machine identity as part of offboarding or workflow change, not leave it running by default. If the automation still needs to exist, reissue it under a fresh ownership decision and a current business purpose.
Technical breakdown
Distributed identity surfaces in human-machine workflows
A distributed identity surface exists when a single business outcome is produced by multiple identity actors rather than one human account. The human may initiate the task, but software agents, embedded automations and assistants can retrieve data, create tickets, route decisions or trigger downstream actions. That breaks the assumption that access can be governed solely at the user record level. The relevant control problem becomes attribution across the chain of action, not just authentication at login. In identity terms, the workflow now contains multiple principals with different privileges and different revocation needs.
Practical implication: Practitioners need to model workflows around actor chains, not single-user sessions, so ownership and audit evidence remain intelligible.
Delegated access versus autonomous action
Delegated action means software is operating on behalf of a human under a bounded purpose. Autonomous action means the system decides when and how to act without a human approval gate at the point of execution. Those are not interchangeable states. Delegated actors still inherit some of the human’s accountability context, but autonomous actors create a separate governance problem because the action can occur after the original human intent is no longer in play. That changes how privilege, logging and revocation work, especially when the machine persists beyond the employee who first enabled it.
Practical implication: Teams should separate delegated automation from autonomous execution in policy, inventory and review processes.
Why identity review fails when machines act inside human workflows
Access review processes assume a stable identity, a stable owner and a stable privilege set over time. Human-machine workflows violate all three. A bot can be created for a temporary purpose, embedded in multiple systems, and still remain active after the original task or employee has changed. Traditional recertification then reviews a credential, not the action path it actually enables. That leaves a governance blind spot where the credential may look valid while the operational use case has drifted. This is why machine identities cannot be treated as lightweight shortcuts.
Practical implication: Governance teams need lifecycle controls that review machine purpose, ownership and runtime usage, not just whether a credential exists.
Threat narrative
Attacker objective: The objective is not necessarily malicious takeover but uncontrolled business action that evades clear attribution and ownership.
- Entry occurs when a human delegates work to an assistant, bot or agent and grants it the access needed to operate across systems.
- Credentialed action follows as the machine identity executes tasks such as triage, routing, code changes or data access inside multiple platforms.
- Escalation happens when the machine keeps operating after the original human context changes, leaving privileges and approvals detached from current accountability.
- Impact is blurred attribution, where security, legal and operational teams cannot reliably prove who initiated the action or owned the outcome.
Breaches seen in the wild
- tj-actions/changed-files compromise 2025: A stolen bot token let attackers poison tj-actions/changed-files so pipelines printed their CI/CD secrets to public logs (CVE-2025-30066).
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
Human-machine identity is now a governance model, not a tooling detail: Once software agents can act alongside people, the enterprise no longer has one accountable identity per worker. That changes the unit of control from the person to the human-plus-machine workflow. IAM and IGA programmes that still anchor decisions to a single user record will miss where accountability actually lives. Practitioners need to treat delegated action as a first-class identity event, not an exception.
Access review assumes identity stability that autonomous and delegated actors do not provide: Recertification works when the privilege holder is stable long enough to be reviewed. A bot or assistant can be created, used and left behind without any obvious leaver event, which means ownership and purpose drift faster than periodic review cycles can detect. The implication is that lifecycle governance has to attach to machine purpose and runtime use, not just account existence.
Accountability collapses when access, intent and execution no longer align: The article’s key insight is that humans think in purpose while machines operate in permission. That gap creates a new kind of governance debt because the identity that authorises an action may not be the one that understands it. This is a direct challenge to privilege assignment, approval design and incident attribution across human, NHI and autonomous layers.
Distributed identity surfaces need a named control concept: identity delegation boundaries: The useful governance question is not simply who logged in, but where delegation ends and substitution begins. That boundary determines whether a machine is acting as an extension of the user or as a separate actor with its own accountability demands. Once organisations cannot draw that line clearly, the whole model of responsibility becomes contestable in audit, incident response and litigation.
The enterprise is moving toward layered identity states, and the control model has to follow: Human, augmented and autonomous states require different logging, ownership and revocation logic. Treating them the same creates false confidence because the same credential can represent very different degrees of agency. The practitioner takeaway is to separate control design by actor state before the first high-impact workflow depends on it.
What this signals
Identity delegation boundaries: The new control point is the line between assistance and substitution. If a workflow cannot show whether a machine is acting on behalf of a person or on its own, accountability will be disputed after the fact. Practitioners should design review, approval and audit models around that boundary rather than around a single account record.
Identity governance now has to account for machine lifecycles inside human work. That means the same discipline used for joiner-mover-leaver processes has to extend to automations, assistants and agents, with ownership and retirement defined at creation rather than discovered during incident response.
For practitioners
- Define identity delegation boundaries Map where a human ends and a machine begins in every workflow that uses assistants, bots or agents. Document which actions are delegated, which are autonomous and which still require explicit human approval.
- Inventory machine identities by business purpose Record the owner, purpose, permissions and retirement condition for each bot, assistant or automation. Treat machine accounts as governed identities, not disposable shortcuts.
- Separate autonomous actions from delegated actions Update policy and logging so autonomous execution is distinguishable from human-directed automation. The governance model should reflect whether the machine is substituting for the user or assisting the user.
- Tie revocation to employee departure and workflow change When the employee leaves or the workflow changes, retire the machine identity as part of the same lifecycle decision. Do not leave credentials in place because the automation still technically works.
- Test incident attribution before deploying agentic workflows Validate that security, legal and audit teams can reconstruct who approved the action, which machine executed it and what authority was used. If the answer is unclear, the governance model is not ready.
Key takeaways
- Human-machine workflows break the old assumption that one user equals one accountable identity, so governance has to move from account-centric control to workflow-centric control.
- The core risk is not that AI is malicious, but that delegated and autonomous actions outlive the human context that authorised them.
- Practitioners should assign ownership, purpose and retirement conditions to machine identities so accountability survives when people and software act together.
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 CSF 2.0 and NIST SP 800-53 Rev 5 set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| OWASP Non-Human Identity Top 10 | NHI-10 — Human Use of NHI | Human workflows now depend on machine identities acting on behalf of people. |
| NHI-01 — Improper Offboarding | The article highlights machine identities that persist after the employee or use case changes. | |
| Recommendation — Separate human-directed automation from autonomous machine action and govern both under explicit ownership. Retire machine identities when the business purpose or human owner changes. | ||
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | The article centres on how permissions and authorisations blur across human-machine workflows. |
| Recommendation — Map delegated and autonomous access paths to PR.AA-05 and verify who can act, not just who signed in. | ||
| NIST SP 800-53 Rev 5 | IA-5 — Authenticator Management | Bots and assistants depend on credentials that must be owned, scoped and retired. |
| Recommendation — Apply IA-5 to manage machine credentials with clear ownership and lifecycle retirement. | ||
| MITRE ATT&CK | TA0004;TA0006 — Privilege Escalation; Credential Access | The article describes machine identities gaining and retaining privileges beyond the intended human context. |
| Recommendation — Track privilege growth and credential persistence when automations inherit access inside workflows. | ||
Key terms
- Human-machine identity surface: The full set of human and non-human actors that can affect a business outcome inside a workflow. It includes people, delegated automations, assistants and autonomous systems, all of which may carry different permissions, owners and revocation rules.
- Delegated action: An action performed by one identity on behalf of another, with some authority borrowed from the principal. In agentic environments, delegated action needs explicit scope, auditability, and revocation boundaries because the executor is not the same as the beneficiary.
- Autonomous Behaviour: Autonomous behaviour is runtime decision-making where the actor can choose actions, sequence, tools, and timing without human approval gates. In identity governance, that changes the control model because access no longer maps neatly to a fixed intended purpose at provisioning time.
- Identity delegation boundary: The point in a workflow where assistance ends and substitution begins. Clear boundaries determine whether a machine is simply extending a human’s work or operating as a separate actor that needs its own ownership, logging and retirement logic.
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.
Published by the NHIMG editorial team on June 10, 2026.
Updated on October 10, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org