Can a model follow a policy announcement without checking its authority?

**Can a model treat a policy announcement as the rule itself?**

Only if the task's authority structure gives that announcement binding status. In a business workflow, a message can summarise a change without carrying every qualification. Draft guidance can also circulate before approval, and an approved policy can remain valid despite an informal suggestion that it will soon change.

**What makes this a useful training problem?**

The model must distinguish confidence of wording from authority. In insurance administration, a team notice might say that all renewals require a new declaration, while the approved instruction limits the requirement to a defined product class. Applying the notice literally would create unnecessary requests and delay work outside that class.

**What could generated material contain?**

A task pack could include an approved policy, a draft amendment, an announcement and a maintained authority register. Their wording can overlap while their status and scope differ. The expected response should identify the controlling source and describe the conflict without silently rewriting either document.

The learning objective is source judgement within an explicit business authority model. It is not a general assumption that formal-looking documents always outrank operational messages.