Commissioned by Kiteworks
Who is responsible when an AI agent acts

Inside the CTRL Vault · not a film still

Who is responsible when an AI agent acts

Who is responsible when an AI agent acts?

Record who authorized the agent and what they permitted. That is a basis for accountability, not a ruling on legal liability.

For day-to-day oversight, responsibility should be clearly assigned to the people who authorize the agent’s work, manage its access and review its actions. Each action should be traceable to the person who approved the workflow and the scope they approved. That establishes accountability; it does not determine legal liability.

An activity dashboard may show what ran without showing who approved it. Record the delegation from the authorizing person to the agent, the permitted task and its data access. Kiteworks’s identity guidance describes workflow credentials that preserve this relationship in access records. Capture attribution when authorization and access happen: shared credentials may identify a system while obscuring the particular agent, workflow or person.

Written ownership should also specify who manages access and reviews incidents. Choose review and intervention checkpoints suited to the workflow’s risks, and preserve the decisions made there. A named authorizer alone does not show that oversight worked, and technical attribution does not settle legal liability.

Put this into practice

  • Record the authorizer and approved scope before work begins.
  • Link each access event to that authorization.
  • Assign incident review roles and retain checkpoint decisions.

Sources

Bring it to a demo

See the controls in your workflow.

Bring one AI task and the data it needs. Ask Kiteworks to show what is allowed, what is denied and the record of each decision.