We rolled out platform tags so organizations can keep a shared catalog of labels in one place and apply those labels to Drive files and folders without turning tags into ad hoc metadata.
What shipped
The new flow splits tags into two parts: organization catalog management in Settings, and object assignment in Drive.
- Admins can create, rename, and remove unused tags from Settings.
- Drive users can assign and remove tags on files and folders they can write.
- Users can filter Drive by one or more tags.
- Shared projects expose host tags read-only to partner viewers.
The model is explicit
Tags are organization-scoped labels with immutable IDs and mutable display names. That gives teams a stable reference for assignments while still letting them rename labels as vocabulary changes.
The tag catalog lives under Settings. The Drive UI reads from that catalog and applies tags to VFS FILE and FOLDER objects. Multiple selected filter tags use AND semantics, so the result set only includes objects that match every active tag filter.
Organization catalog
|
v
Drive object assignment
|
v
Visible chips + tag filtersWhat tags are not
Tags are advisory labels. They do not replace authorization, policy enforcement, or legal hold. They also are not stored as the source of truth in object metadata_json.
That distinction matters. Visibility is useful, but enforcement still belongs in the policy layer. Tags help people classify and find content. They do not decide who may access it.
Feature gating and tenant boundaries
Platform tags are gated per tenant. When the feature is off, mutations fail closed. Existing assignments may remain visible in read-only mode, but new tag catalog changes and new assignments are blocked.
Host organizations own the catalog on shared projects. Partner viewers can see those tags, but they cannot assign, remove, or manage the catalog.
Why this matters
Teams need a shared vocabulary before they can classify data consistently. Without a catalog, tagging turns into local conventions and one-off labels. With a catalog, the system can enforce naming rules, limits, and audit events around the same set of labels.
The rollout records tag creation, updates, deletions, assignments, and removals. That gives operators a traceable record of how classification changes over time.
The practical outcome
The result is a smaller, clearer path for classification in Drive: one catalog, one assignment model, one filter surface, and explicit tenant boundaries. It keeps labels useful without pretending they are controls.