Control · C-107 · Security & Resilience

Authority check at the point of execution

Immediately before an agent's action takes effect, its authority is confirmed again: the delegation, approval, credential and business conditions must all still hold, or the action does not proceed.

Control type
Preventive
Risks addressed
5
Audience
Heads-of · Practitioner
Frameworks
Article 14 · AML.M0026 · ASI03
Reviewed
2026-11-17

What it does

It closes the gap between the moment an action is authorised and the moment it takes effect. The component that carries out the action checks again, at that time, that everything which justified it still holds, including after a delay, a retry or a hand-off to another agent. If any of it has lapsed or been withdrawn, the action stops. Identity and access management owns the mechanism; the business owner of each process decides which conditions must still hold.

How to implement it

  • Bind each approval and delegation to the exact action and parameters, with an expiry.
  • Check authority, approval and conditions again at the point of effect, not only when the task starts.
  • Repeat the check on every retry, queued action and alternative execution path.
  • Make the check and the action a single step where the system allows, so authority cannot change between them.
  • Carry a user's cancellation to every queued or delegated part of the task.
  • Fail closed when the check cannot be completed.

Evidence it operates

Design: for each consequential operation, a statement of what must hold at execution and where it is checked. Operation: a test in which authority is changed after a task starts, by withdrawing an approval, narrowing a delegation, revoking a credential or cancelling the request, and the action does not take effect on any path, retries included.

Where it fits

Built in at design and operating on every consequential action. It is the primary control for action on stale or withdrawn authority. It turns least-privilege agent identity and approval gates from checks made once into checks that hold until the action is done.

Limits

It confirms that authority still holds; it cannot judge whether a permitted action is right in its context. Each check adds delay and depends on the systems that hold the conditions being available, so a choice has to be made between waiting and refusing. It cannot recall an effect already in the world.

Risks this mitigates

Primary mappings

Action on stale or withdrawn authority High
Checks authority, approval and conditions again at the point of effect. An agent's action takes effect after the approval, delegation or condition that justified it has expired, been withdrawn or ceased to hold, so the effect occurs without valid authority.

Supporting mappings

Agent identity and privilege abuse High
Stops a credential or delegation being used after it should have lapsed. An agent inherits, shares or escalates credentials and delegated authority beyond what its task needs, so a compromise or error acts with far wider access than intended.
Inability to stop or roll back AI High
Makes a decision to stop reach actions already queued or being retried. A misbehaving system cannot be disabled, reverted to a known good version, or have its actions reversed quickly enough to limit harm.
Privilege escalation through agent delegation chains High
Confirms when the action takes effect that the delegated authority is still valid. An agent passes work to other agents with authority that is wider, longer-lasting or deeper than its own, so a delegated task can do what no agent in the chain was granted.
Unauthorised commitments by agents High
Confirms authority to commit at the moment of commitment. An agent makes a statement, offer or transaction that binds the organisation, without the authority a person in the same position would need.

Framework and clause references

FrameworkClauseTitle
EU AI ActArticle 14Human oversight Makes an oversight decision to stop effective on queued and retried actions.
MITRE ATLASAML.M0026Manage AI Agent Service Access Managing what an agent may access, applied at the point of execution.
OWASP Top 10 for Agentic ApplicationsASI03Identity and Privilege Abuse Confirms that authority is still valid when used.

Related terms

How to cite

Axiom Verity (2026). Authority check at the point of execution. Reference Library, C-107. https://axiom-verity.com/controls/authority-check-at-execution. Reviewed 2026-11-17.