Databricks Generative AI Engineer Associate: Governance Practice Questions

In Unity Catalog, the SELECT privilege on a Vector Search index grants the ability to query (read) the index, which is exactly what a serving endpoint needs during inference. This follows the principle of least privilege. Option A (USE CATALOG alone) is a prerequisite for navigating the catalog hierarchy but does not itself grant read access to the index data. Option C (ALL PRIVILEGES on the schema) would grant far more than necessary, including the ability to create, modify, and delete objects — violating least privilege. Option B (CREATE TABLE) grants the ability to create new tables in the schema, which is unrelated to reading the index and exceeds what is needed.

Option A is correct: a PII detection/masking step applied to retrieved chunks before prompt construction ensures that even if a document contains raw PII, that PII is replaced or redacted before it enters the LLM context. Option C is correct: Unity Catalog dynamic data masking on the source table prevents PII from being returned to any downstream process (including the retrieval step) for users or service principals that do not have explicit access, stopping PII at the data layer. Option D (time travel) is an auditing/versioning feature and does not mask PII. Option B (max_tokens) limits response length but does not prevent PII from being included in whatever tokens are generated. Option E (Secret Scopes) is for managing API keys and passwords, not for masking column-level PII in Delta tables.

Option B is correct because separating sensitive documents into a Unity Catalog schema with restricted permissions enforces access control at the data platform level rather than relying on application logic alone. Having a dedicated index ensures that unauthorized users' retrieval queries never touch restricted content, even if the application layer is misconfigured. Option A is wrong in practice because Databricks Vector Search indexes must be backed by a Delta table or materialized view, not a dynamic (logical) view; the row-level filtering would not propagate into the vector index at query time. Option C is wrong because an application-layer metadata filter is a 'soft' control that is easily bypassed by a misconfigured or malicious caller; platform-level enforcement is preferred. Option D is wrong because the LLM itself does not perform decryption; data is decrypted before being passed to the model, so this approach does not prevent the restricted content from appearing in retrieved chunks passed to the LLM prompt.

Option F is correct because EXECUTE on the model object is the specific privilege needed to load and serve a registered model. Option D is correct because USE CATALOG and USE SCHEMA are prerequisite hierarchical privileges required to access any object within a catalog/schema; without them, even object-level grants are ineffective. Option E is correct because if model binaries are stored in a Unity Catalog Volume, the service principal also needs READ VOLUME to access the artifact files at serving time. Option C is wrong because the trace data is accessed through MLflow's UI and API rather than a direct SELECT on a Delta table; this is not the standard access pattern. Option B is wrong because ALL PRIVILEGES violates least-privilege and grants unnecessary permissions (e.g., CREATE, MODIFY) beyond what serving requires. Option A is wrong because MODIFY allows writing/editing the model object, which is broader than what data scientists need merely to read trace metadata.

Option B (EXECUTE) is the correct Unity Catalog privilege that allows a principal to load and use a registered model (including calling model serving endpoints backed by it) without the ability to register new versions. Option D (SELECT) is a privilege applied to tables and views, not to model objects. Option C (APPLY TAG) only permits adding metadata tags to the object and does not enable model loading. Option A (READ PRIVATE FILES) relates to external locations and storage credentials, not to model registry objects.