SnowPro Specialty: Gen AI · 18% of the exam

Snowflake Gen AI Governance: free practice questions

5 sample questions from our 46-question bank for this domain — answers and explanations included. These are the same scenario-based style as the real Snowflake exam.

1. Source documents contain PII columns that some users must not see, even when those columns feed a RAG pipeline. Which existing Snowflake governance controls should be applied to the source and embedding tables?

  • A. Masking policies and row access policies enforced through RBAC✓ Correct
  • B. Disabling Time Travel
  • C. Increasing warehouse size
  • D. Turning off result caching
Explanation

Cortex pipelines inherit Snowflake's governance: dynamic data masking and row access policies, enforced via RBAC, restrict which users see sensitive columns/rows in the source and embedding tables. The other options don't control data access.

2. Which combination of controls would a well-governed Cortex deployment for a regulated industry most plausibly rely on together?

  • A. SNOWFLAKE.CORTEX_USER for access, CORTEX_MODELS_ALLOWLIST for approved models, CORTEX_ENABLED_CROSS_REGION for residency, Cortex Guard for content safety, and existing RBAC/masking/row access policies on the data itself✓ Correct
  • B. Only CORTEX_MODELS_ALLOWLIST, since it alone covers access, residency, and content safety
  • C. Only resource monitors, since cost control is the sole governance concern in regulated industries
  • D. Only Cortex Guard, since content safety is a superset of the other governance concerns
Explanation

A comprehensive governance posture layers all of these distinct, complementary controls together, each addressing a different risk. The allowlist alone does not cover access, residency, and content safety — those are each separate controls. Cost control via resource monitors is one important piece but is not the sole governance concern, especially for regulated data. Cortex Guard addresses content safety specifically; it is not a superset that also covers access, model approval, or residency.

3. A team building a customer-facing chatbot on Cortex Agents wants to make sure the chatbot never surfaces another customer's masked PII, even though the chatbot's role has broad table SELECT for operational reasons. What should they rely on?

  • A. The existing masking policy on the PII column, which applies to the chatbot's role's reads exactly as it would for a person running the same query✓ Correct
  • B. A brand-new AI-specific masking mechanism that must be created separately from the standard masking policy framework
  • C. Disabling SNOWFLAKE.CORTEX_USER for the chatbot's role, which would also disable its ability to answer any question
  • D. Trusting that Cortex Agents automatically infer and redact PII without any policy configuration
Explanation

Standard masking policies apply to any reader, including a Cortex Agent's underlying queries — no separate mechanism is needed. There's no AI-specific masking framework described; the existing policy framework covers this. Disabling CORTEX_USER would break the chatbot's core function entirely, an overcorrection that also fails the goal of it still answering safely. Cortex Agents do not automatically infer and redact PII on their own; that protection comes from configured masking policies on the underlying data.

4. An organization must ensure Cortex inference for a workload only runs in regions where their data is allowed to reside. Which account setting governs whether requests can be served from another region?

  • A. CORTEX_ENABLED_CROSS_REGION✓ Correct
  • B. AUTOCOMMIT
  • C. STATEMENT_TIMEOUT_IN_SECONDS
  • D. QUERY_TAG
Explanation

CORTEX_ENABLED_CROSS_REGION controls whether Cortex may process a request in a different region when the model isn't available locally. Setting it appropriately enforces data-residency requirements. The other parameters are unrelated to cross-region inference.

5. A team wants to revoke a role's ability to call Cortex functions without touching any of that role's existing table-level grants. What is the correct action?

  • A. Revoke the SNOWFLAKE.CORTEX_USER database role from that role✓ Correct
  • B. Revoke SELECT on every table the role can access
  • C. Disable the account's virtual warehouses
  • D. Set CORTEX_MODELS_ALLOWLIST to an empty list for the entire account
Explanation

Revoking SNOWFLAKE.CORTEX_USER is the direct, additive way to cut off Cortex access without altering table grants. Revoking SELECT on every table would remove far more than Cortex access and would break other unrelated workloads. Disabling warehouses would stop essentially all query activity, not just Cortex calls, and is a disproportionate action. Emptying the allowlist for the whole account would affect every role's Cortex access, not just the target role, which is broader than intended.

41 more questions in this domain

Practice the full bank with instant grading, flashcards, and a timed mock exam.

Start practicing free