/
New feature Public Preview

Role-based access control (RBAC)

Role-based access control (RBAC) is a Databricks data governance capability, introduced July 2026.

Databricks access-control model where a user assumes a role - implemented as a group - and acts with only that role's permissions for the duration of the session, instead of the permissions they have accumulated.

  • A 'role' here is not a new kind of object at all - it's an ordinary Databricks group that someone has been granted the Assume permission on, wearing a different hat.
  • Normally your permissions pile up from every group you belong to and follow you everywhere, so assuming a role is one of the rare logins that can leave you seeing less rather than more.
  • Anything you make while acting as a role - notebooks, queries, SQL warehouses, jobs - is owned by the role rather than by you.

Limitations: A user can act as only one role at a time. When acting as a role, creating Agent Bricks agents, managing alerts, and Lakeflow pipelines are unsupported, and creating Vector Search indexes fails (create as a user, then reassign ownership to the role); the workspace SCIM API cannot create or manage groups, and roles still inherit whatever the workspace users group can see.

Open in REbricked →
Category
Data governance
Introduced
July 2026
Also known as
RBAC, role-based access control, exclusive access, role switcher
Verified
2026-08-03

Sources

Related in Data governance