We answer the model-handling and data-flow questions third-party risk reviewers most often ask. Briefs, documents, modules, decks, and learner data are never used for model training — by us or by our sub-processors.