AI termBrowse the neighboring terms

Agent Ops / Industry term

Privacy label

Metadata that classifies a record's sensitivity, handling requirements, permitted uses, or disclosure audience.

A label might indicate public, internal, confidential, regulated, or a project-specific category. The label is input to policy; it does not enforce the policy by itself, and one tag may be too coarse for purpose, jurisdiction, retention, or field-level restrictions. Defaults should follow the system's actual risk and data contract rather than always mean local-only.

Builder example

Enforcement belongs in access control, retrieval filtering, egress policy, logging, and retention systems. A model instruction to honor a label is helpful context but not a security boundary. Derived artifacts need labels and provenance that account for all contributing sources.

Common confusion: A privacy label controls what the assistant may do with a record, while a permission scope controls which account the assistant may reach into at all. Scope decides whether it can open your email; the label on a given message decides whether its contents may travel into an outgoing draft.