/
New feature GA

Attribute-based access control (ABAC)

Attribute-based access control (ABAC) is a Databricks data governance capability, introduced June 2025.

Unity Catalog access-control model that grants or restricts access by evaluating governed tags (attributes) on securables, so one policy on a catalog or schema applies to everything beneath it.

  • Tag a column 'pii' once and a single catalog-level policy masks it everywhere beneath - no more re-granting access table by table.
  • The same tag-driven policies do both row filtering and column masking, on tables, materialized views, and streaming tables alike.
  • Its GRANT policies hand out privileges to whatever matches a tag condition instead of object by object, and in August 2026 that grew from models alone to model services, MCP services, and agent services.

Limitations: Needs serverless compute or Databricks Runtime 16.4 and above, and only governed tags count; policies can't be applied directly to views, though they are still enforced when a view reads a protected table; quotas apply (10,000 policies per metastore, 100 per catalog or schema, 50 per table, 20 principals per policy); time travel queries and deep or shallow clones fail on policy-protected tables unless the principal is exempted; and AI Search indexes don't enforce row filters or column masks.

Open in REbricked →
Category
Data governance
Introduced
June 2025
Announced at
GA announcement blog, May 2026
Also known as
ABAC, attribute-based access control
Verified
2026-08-20

Sources

Related in Data governance

Guides that mention this