Data catalog

Central inventory. Metadata: source, owner, sensitivity, retention, allowed uses. Tools: Collibra, Alation, Purview.

Advertisement

Lineage tracking

Source → transformation → dataset → model. On personal data request, identify all downstream use.

Advertisement

Classification

Public / Internal / Confidential / Restricted. Different handling per class. AI training rules per class.

Retention

Automated deletion per policy. Not just DBs — training data, embeddings, model checkpoints.