A course invites you to control AI but doesn't show you how to do it. I propose exercises on sources, limit cases and verification of data.
In the example, participants must explain why they accept or reject an output. Useful doubt is not generalized distrust.
It is a set of questions relevant to the task. Training should make it possible to recognise when another competence is needed, avoiding the obligation of supervision being assigned to persons without preparation and possibility of updating.