0% complete

Governance, safety, and security · Lesson 1 of 3

Unity Catalog and AI governance

Apply governed access and traceability to data and AI assets.

Lesson

AI applications inherit the risks of the data and assets they can access. Unity Catalog can provide centralized governance for supported data and AI assets, including permissions and discoverability. Design permissions around least privilege: application identities should see only the sources and operations required for the task. Verify the effective identity and access path in the deployed environment, not only in a developer notebook. Governance also includes lineage, ownership, approval, and lifecycle decisions. A user asking a question should not receive material they cannot otherwise access simply because a shared index returned it. Apply authorization-aware filtering before context reaches the model, and test that different identities receive only permitted records. Start governance design with identities and data boundaries. Identify the human user, application service principal, governed source, index, endpoint, and permissions needed for each request. Apply access filters before private records are assembled into model context; hiding a citation after generation does not undo disclosure. Keep ownership and lineage discoverable so a source update or access revocation can be followed through derived assets. Verify the deployed application identity in the target workspace, since a developer’s notebook privileges can be much broader than production permissions. Implementation drill: trace one representative request through the stages in the diagram. Write down what enters each stage, what output or evidence it produces, and which identity is allowed to perform it. Then test an empty input, an unauthorized request, a dependency timeout, and an unexpected result. For each case, state whether the application should retry, reject, abstain, or ask a person for help. Keep the observed failure as a regression example so changes to prompts, models, data, or code can be checked against the same behavior. Practice task: When should authorization be applied to retrieved private documents?

Visual guide

Apply governed access to retrieved data

  1. Step 01 Identify

    Determine user and service identities.

  2. Step 02 Authorize

    Check access to source records.

  3. Step 03 Retrieve

    Fetch only permitted evidence.

  4. Step 04 Audit

    Retain governed lineage and ownership.

Key points

  • Use least-privilege identities and governed access to data and AI assets.
  • Apply user-appropriate authorization before private data enters a prompt.
  • Test data access using the deployed application identity.

Check your understanding

Practice

When should authorization be applied to retrieved private documents?

Your learning progress

Finish this lesson