To read a Unity Catalog table, a user needs USE CATALOG, USE SCHEMA and SELECT; row filters, column masks and tag-based policies then narrow what comes back. External engines get short-lived storage credentials, and filtered tables reach them through cross-engine ABAC, GA since October 1, 2026. Atlan brings Unity Catalog’s tags, lineage and metric views into one graph.
Sources: Databricks, “Unity Catalog permissions model concepts”, “Unity Catalog credential vending for external system access” and “Cross-engine attribute-based access controls (ABAC)”; release notes, “October 2026”; Atlan, What does Atlan crawl from Databricks?.
Atlan, the context layer for AI, needs one connection per metastore, covering what its service principal can see (Unity Catalog governance and connections). It pulls tags, lineage and usage into the Enterprise Data Graph and syncs tag updates back. Enforcement stays in Unity Catalog.
| Question | Answer |
|---|---|
| What does a table read need? | USE CATALOG, USE SCHEMA and SELECT |
| How do external engines reach files? | Vended credentials, one hour by default for Iceberg clients |
| Filtered tables from outside Databricks? | Through cross-engine ABAC (GA October 1, 2026), on a managed table with catalog commits |
| Open source? | An Apache 2.0 project; v0.7.0 released October 13, 2026 |
Sources: Databricks, “Access Databricks tables from Apache Iceberg clients”; release notes, “October 2026”; unitycatalog/unitycatalog, GitHub.
How does Unity Catalog organize data and metadata?
Unity Catalog stores metadata in a metastore, and “each Databricks region requires its own Unity Catalog metastore” (Databricks, “Create a Unity Catalog metastore”). Tables, views, volumes, functions, models and services follow a three-level namespace, catalog.schema.object, while storage credentials, external locations, connections and shares sit directly under the metastore (Databricks, “What is Unity Catalog?”). Atlan’s pages on Databricks metadata management and metadata management for AI go deeper here, and data lakehouse for AI explains why lakehouse storage needs a metadata layer.
Two objects connect Unity Catalog to cloud storage. A storage credential, such as an IAM role, “represents an authentication and authorization mechanism for accessing data stored on your cloud tenant”, and an external location pairs a storage path with that credential (Databricks, “Connect to cloud object storage using Unity Catalog”). Managed tables and volumes give Unity Catalog both governance and the file lifecycle; external ones get governance only. A managed table’s files always stay in your cloud account; drop the table and they are permanently deleted after a recovery period, 7 days by default (Databricks, “Managed versus external assets in Unity Catalog”).
Catalogs are open to every workspace on the metastore until you bind them to named workspaces; then “access from an unbound workspace is denied, even for users with explicit privilege grants on the catalog” (Databricks, “Workspace-catalog binding”). The metastore holds names and privileges, and for managed assets Unity Catalog also controls the files; reads from other engines follow different rules.
How does Unity Catalog decide whether a query can read a table?
Databricks documents a governed read in two stages: Unity Catalog evaluates privileges and policies, then the Databricks Runtime enforces them. ABAC denies access when Databricks cannot verify security.
How a read on Databricks compute is authorized:
- The name resolves as
catalog.schema.objectin the region’s metastore. - Unity Catalog checks USE CATALOG, USE SCHEMA and the operation privilege, such as SELECT, granted directly or inherited from a catalog or schema.
- For ABAC, Unity Catalog finds the policies whose scope covers the table, checks the user against each policy and its conditions, and sends the effective row filter or column mask to the Databricks Runtime.
- The Databricks Runtime enforces the filter and masks during query execution.
- If Databricks cannot verify that it can enforce every applicable ABAC policy, for example on a runtime below 16.4, access is denied.
Unity Catalog also records lineage and an audit event for the query. External engines take a different path, covered below.
Sources: Databricks, “Unity Catalog permissions model concepts” and “Row filter and column mask policy evaluation and runtime behavior”.
“Having only the SELECT privilege on a table is not sufficient to read it” (Databricks, “Unity Catalog permissions model concepts”). Grant usage on the containers to a group; SELECT on a schema or catalog reaches its current and future tables. Access control for AI agents covers the agent identity side. Syntax per Databricks, “Manage privileges in Unity Catalog” and the SQL reference “GRANT”:
GRANT USE CATALOG ON CATALOG main TO `finance-team`;
GRANT USE SCHEMA ON SCHEMA main.default TO `finance-team`;
GRANT SELECT ON TABLE main.default.orders TO `finance-team`;
How do row filters, masks and ABAC narrow a read?
A row filter is a SQL UDF evaluated per row at query time, and only one distinct row filter can resolve per user and table (Databricks, “Row filters and column masks”). ABAC attaches policies on governed tags at the catalog, schema or table level, or at the metastore level in Beta (Databricks, “Attribute-based access control in Unity Catalog”). Role-based access control in a context platform and Databricks data governance show how agents inherit these rules.
Compute access mode also matters. Dedicated access compute cannot read a table with row filters or column masks on Runtime 15.3 or below; on 15.4 LTS or above it can, through serverless compute, if the workspace has serverless enabled (Databricks, “Row filters and column masks”).
What breaks a governed query, or changes it silently:
| What happens | Result |
|---|---|
| Filter UDF parameter type differs from the column type, ANSI mode off | Values silently become NULL with no error, and a row filter can return all rows |
| A governed tag a policy references is deleted | Queries in scope fail with UC_ABAC_UNKNOWN_TAG_POLICY |
| Compute below Runtime 16.4 reads an ABAC table | Access denied |
You read system.query.history |
statement_text returns <REDACTED> by default since August 26, 2026, except for account admins and the databricks_pii_access group |
Sources: Databricks, “Row filters and column masks” (type mismatch) and “Row filter and column mask policy evaluation and runtime behavior” (tag deletion, Runtime 16.4); release notes, “August 2026”.
Outside Databricks compute this changes: a vended credential carries the configuring principal’s privileges, not the filters and masks.
How do engines outside Databricks reach Unity Catalog data?
External engines read with credentials Unity Catalog vends over the Iceberg REST catalog or Unity REST APIs. For them, row filters and masks hold only through cross-engine ABAC, which filters the data on Databricks serverless compute.
“Credential vending grants short-lived credentials using the Unity Catalog REST API,” and the credentials “inherit the privileges of the Databricks principal used to configure the integration” (Databricks, “Unity Catalog credential vending for external system access”). The default expiry is one hour (Databricks, “Access Databricks tables from Apache Iceberg clients”). The caller needs EXTERNAL USE SCHEMA (Public Preview), which schema owners lack by default (Databricks, “Unity Catalog privileges reference” and “Unity Catalog permissions model concepts”). Table formats are covered in Apache Iceberg on Databricks.
Standard vending stops at filtered tables: “You cannot use Iceberg REST catalog or Unity REST APIs to access tables with row filters or column masks” (Databricks, “Row filters and column masks”). Cross-engine ABAC, GA October 1, 2026, needs a managed table with catalog commits and bills the serverless compute it runs on. Its documented clients are Apache Spark, with Delta-Spark 4.1 or Iceberg-Spark 1.11 and above, and “only reads are supported from external engines” under these controls (Databricks, “Cross-engine attribute-based access controls (ABAC)”).
Who is reading decides what Unity Catalog can still enforce:
| Reader | Grants | Filters, masks, ABAC |
|---|---|---|
| Databricks compute (ABAC: Runtime 16.4+ or serverless) | Checked | Enforced |
| Dedicated access compute, Runtime 15.4 LTS+, serverless enabled | Checked | Enforced on serverless compute |
| Dedicated access compute, Runtime 15.3 or below | Checked | Filtered tables cannot be read |
| External engine, unfiltered table | Checked, plus EXTERNAL USE SCHEMA | Not applicable |
| External engine, filtered table, standard credential vending | Checked | Use cross-engine ABAC (standard vending does not cover filtered tables) |
| External Apache Spark client through cross-engine ABAC | Checked | Enforced on billed serverless compute, reads only |
| OpenSharing recipient | Share grant | Recipient can apply its own filters and masks to shared tables; providers can share views over ABAC-protected tables |
| Direct cloud-storage access to an external table | Bypassed | Bypassed |
Delta Sharing became OpenSharing on June 10, 2026 (Databricks release notes, “June 2026”), and “the provider’s ABAC policy doesn’t govern the recipient’s access.” Share owners can also share views that reference ABAC-protected base tables (Databricks, “OpenSharing and ABAC”). Direct external access to an external table “bypasses Unity Catalog access control, auditing, and lineage” (Databricks, “Unity Catalog best practices”). See Unity Catalog limitations, and how agents reach Unity Catalog data over MCP.
What does Unity Catalog record about each read?
Unity Catalog records lineage automatically for queries run on Databricks; system tables keep a rolling 1-year window and Catalog Explorer keeps it indefinitely (Databricks, “Lineage in Unity Catalog”). External lineage reached GA on June 8, 2026 (Databricks release notes, “June 2026”). Registered external objects appear in the lineage graph view, but external lineage “is not recorded in the lineage system tables” (Databricks, “External lineage”). Audit events land in system.access.audit, labeled Public Preview, with 365 days of free retention (Databricks, “System tables reference”).
Automatic lineage covers queries run on Databricks, and Lakeflow Connect managed ingestion records source lineage (Databricks release notes, “June 2026”); for other systems, you register external lineage. Four Atlan pages follow the lineage path agents need: data lineage for AI, AI-ready data lineage, lineage over MCP and Databricks data lineage.
Does Unity Catalog store business definitions?
Yes. Metric views are “the core implementation of Unity Catalog semantics” (Databricks, “Unity Catalog metric views”), and Business Semantics reached GA on April 2, 2026 (Databricks blog, “Announcing General Availability and Open Sourcing of Unity Catalog Business Semantics”). The Discover page and domains reached GA on October 8, 2026, and Pages, for governed definitions of business terms and KPIs, entered Beta on August 12, 2026 (Databricks release notes, “October 2026”, “August 2026”).
Databricks documents querying metric views from external BI tools (Databricks, “Unity Catalog metric views”). See what a semantic layer is, then Unity Catalog metric views, Unity Catalog’s semantic layer, glossary and domain feature status, and how Genie Ontology reads them.
What changed in Unity Catalog in the last 12 months?
In twelve months ABAC went from Public Preview to GA, and external engines gained a path to filtered tables.
| Date | Change |
|---|---|
| Nov 3, 2025 | ABAC Public Preview |
| Apr 2 to 28, 2026 | GA: governed tags, Business Semantics, ABAC |
| May 21 to Jun 10, 2026 | GA: managed Iceberg tables, external lineage; Delta Sharing renamed OpenSharing |
| Aug 3 to 26, 2026 | MANAGE no longer needs USE CATALOG or USE SCHEMA on the object where it is granted; Pages Beta; query text masked by default |
| Sep 8 to 17, 2026 | Beta: ABAC DENY policies, metastore-level ABAC |
| Oct 1 to 12, 2026 | GA: cross-engine ABAC, Discover page and domains; tag propagation Beta |
Sources: Databricks platform release notes, November 2025 to October 2026; Databricks blog, April 2, 2026.
On June 16, 2026 Databricks announced cross-region governance as “coming soon in preview” (Databricks blog, “What’s new with Unity Catalog at Data + AI Summit 2026”); no release note through November 2, 2026 shows it shipped. It is one of the Data + AI Summit 2026 announcements.
What should you decide when you set up Unity Catalog?
Five decisions shape a Unity Catalog setup.
- Check enablement. Workspaces created after November 8, 2023 already have Unity Catalog (Databricks, “What is Unity Catalog?”).
- Confirm the metastore. “You can have only one metastore per region” (Databricks, “Unity Catalog best practices”).
- Choose managed or external for each table and volume, and create the storage credentials and external locations that external data needs.
- Grant to groups at the container level, and bind each catalog to the workspaces that should reach it.
- Decide external access early, using the checklist below.
The Databricks data catalog setup guide covers connecting Databricks to an external data catalog. Step 5 decides what engines outside Databricks can read; for agents inside it, such as Databricks AI/BI Genie, see the context Genie needs to answer correctly, and for the wider picture, AI-ready data.
Which reads leave Unity Catalog’s enforcement path, and what breaks inside it? A checklist
- RP-01. List every principal holding EXTERNAL USE SCHEMA.
- RP-02. For each filtered table they reach, plan cross-engine ABAC (Apache Spark clients, reads only) or accept no external read.
- RP-03. Budget the serverless compute cross-engine ABAC runs on.
- RP-04. For each OpenSharing share, decide whether to share views over ABAC-protected tables or rely on the recipient’s own filters and masks.
- RP-05. Find external tables other cloud identities can reach.
- RP-06. Confirm compute touching ABAC tables runs Runtime 16.4+ or serverless.
- RP-07. Before deleting a governed tag, find the policies that reference it.
- RP-08. Match row-filter UDF parameter types to column types.
- RP-09. Register external lineage for outside systems.
- RP-10. Plan for the 1-year window in the lineage system tables (Catalog Explorer keeps lineage indefinitely) and for redacted query text.
- RP-11. List the definitions agents outside Databricks need.
What does Unity Catalog leave to the rest of your estate?
Inside Databricks, Unity Catalog holds grants, policies, lineage and metric definitions, and it enforces filters and masks only where Databricks compute, including the serverless layer behind cross-engine ABAC, runs the read. A vended storage credential carries file access. Agents working across the estate also need metric definitions, owners and lineage from every other system they touch.
Atlan puts the Databricks side in one graph with those systems, as a context layer for Databricks. It maps metric views to a DatabricksMetricView asset (What does Atlan crawl from Databricks?), and reverse sync updates existing Databricks tags without creating or deleting any (Manage Databricks tags). Lineage comes from Unity Catalog and usage from system tables (Extract lineage and usage from Databricks). Grants, row filters, column masks and audit logs “aren’t imported into Atlan” (Unity Catalog governance and connections).
Atlan MCP is “a hosted server, available on every Atlan tenant with nothing to install,” and every action respects permissions set in Atlan (Atlan MCP). That is how an enterprise context layer reaches agents (why agents need one). For Genie, see Genie Ontology and the Atlan context layer and adding enterprise context to a Genie agent. For metrics that span platforms, see Unity Catalog versus a cross-platform semantic layer and context engineering.
Start with RP-01. The principals holding EXTERNAL USE SCHEMA are your first map of who can take vended reads outside Databricks compute.
FAQs about Databricks Unity Catalog
1. Is the SELECT privilege enough to read a Unity Catalog table?
No. A table read needs USE CATALOG on the catalog, USE SCHEMA on the schema and SELECT on the table. A privilege granted on a catalog or schema applies to its current and future child objects.
2. Can engines outside Databricks read Unity Catalog tables?
Yes, through the Iceberg REST catalog or Unity REST APIs with short-lived vended credentials. Filtered or masked tables need cross-engine ABAC, GA since October 1, 2026, on a managed table with catalog commits.
3. Is Unity Catalog open source?
An open-source implementation exists under Apache 2.0, as a sandbox-stage project of the LF AI & Data Foundation; version 0.7.0 shipped October 13, 2026 (unitycatalog/unitycatalog, GitHub; “Unity Catalog – LFAI & Data”).
4. Does Unity Catalog track lineage automatically?
Yes, for queries run on Databricks, down to the column level. For outside systems you register external lineage, GA since June 8, 2026, which appears in the lineage graph but not in the lineage system tables.
5. What does Atlan import from Unity Catalog, and what stays there?
Atlan imports tags, lineage, usage and metric views, and can push tag updates back to existing Databricks tags. Grants, row filters, column masks and audit logs are enforced inside Unity Catalog and aren’t imported into Atlan.
Sources
- “What is Unity Catalog?”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/, September 11, 2026 - “Unity Catalog permissions model concepts”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/access-control/permissions-concepts, September 11, 2026 - “Unity Catalog privileges reference”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/manage-privileges/privileges, November 2, 2026 - “Manage privileges in Unity Catalog”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/manage-privileges/, October 15, 2026 - “GRANT”, Databricks SQL language reference,
docs.databricks.com/aws/en/sql/language-manual/security-grant, September 11, 2026 - “Create a Unity Catalog metastore”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/create-metastore, October 14, 2026 - “Unity Catalog best practices”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/best-practices, October 30, 2026 - “Connect to cloud object storage using Unity Catalog”, Databricks,
docs.databricks.com/aws/en/connect/unity-catalog/cloud-storage/, September 11, 2026 - “Managed versus external assets in Unity Catalog”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/managed-versus-external, October 30, 2026 - “Workspace-catalog binding”, Databricks,
docs.databricks.com/aws/en/catalogs/binding, September 11, 2026 - “Row filters and column masks”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/filters-and-masks/, October 9, 2026 - “Attribute-based access control in Unity Catalog”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/abac/, October 22, 2026 - “Row filter and column mask policy evaluation and runtime behavior”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/abac/policy-evaluation, October 22, 2026 - “OpenSharing and ABAC”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/abac/opensharing, October 9, 2026 - “Unity Catalog credential vending for external system access”, Databricks,
docs.databricks.com/aws/en/external-access/credential-vending, October 30, 2026 - “Access Databricks tables from Apache Iceberg clients”, Databricks,
docs.databricks.com/aws/en/external-access/iceberg, October 27, 2026 - “Cross-engine attribute-based access controls (ABAC)”, Databricks,
docs.databricks.com/aws/en/external-access/cross-engine-abac, October 20, 2026 - “Lineage in Unity Catalog”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/data-lineage, October 29, 2026 - “External lineage”, Databricks,
docs.databricks.com/aws/en/data-governance/unity-catalog/external-lineage, October 7, 2026 - “System tables reference”, Databricks,
docs.databricks.com/aws/en/admin/system-tables/, October 28, 2026 - “Unity Catalog metric views”, Databricks,
docs.databricks.com/aws/en/uc-semantics/metric-views/, October 29, 2026 - Databricks platform release notes, “November 2025” to “October 2026”,
docs.databricks.com/aws/en/release-notes/product/ - “Announcing General Availability and Open Sourcing of Unity Catalog Business Semantics”, Databricks blog,
databricks.com/blog/redefining-semantics-data-layer-future-bi-and-ai, April 2, 2026 - “What’s new with Unity Catalog at Data + AI Summit 2026”, Databricks blog,
databricks.com/blog/whats-new-unity-catalog-data-ai-summit-2026, June 16, 2026 - GitHub - unitycatalog/unitycatalog: Open, Multi-modal Catalog for Data & AI · GitHub
- Unity Catalog – LFAI & Data, LF AI & Data Foundation
- Unity Catalog governance and connections, Atlan Documentation
- What does Atlan crawl from Databricks?, Atlan Documentation
- Manage Databricks tags, Atlan Documentation
- Extract lineage and usage from Databricks, Atlan Documentation
- Atlan MCP, Atlan Documentation