Should the AI see everything a user is entitled to?

The instinct is to give the model everything the requester could access and let relevance sort it out. That maximizes recall and simultaneously maximizes what a single injection or ranking error can surface, which is a poor trade in regulated work.

Taxonomy scoping in Centralpoint constrains a prompt to a branch of the hierarchy, so a benefits question draws from benefits content even where the requester is entitled to more. Scope boundaries in the governance tier reinforce it, and because both are properties of the index rather than post-filters, the narrowing is structural rather than advisory.