SnowPro Advanced: Security Engineer Question 426
Single answerUse Cortex Analyst to support secure data exploration:A financial services company wants business analysts to explore account activity using Cortex Analyst with natural-language questions, but the security team must ensure analysts only see rows and columns they are already authorized to access. The data is stored in Snowflake tables protected with existing row access policies and masking policies. Which approach best supports secure data exploration while preserving those controls?
- A
Create a semantic model over secure views or tables that already have the required row access and masking policies, and ensure analysts query Cortex Analyst using their own Snowflake identities/roles.
- B
Export the governed data to a separate reporting schema without policies, because Cortex Analyst applies its own security layer independently of Snowflake object policies.
- C
Grant analysts a high-privilege role for Cortex Analyst sessions so the model can answer all questions completely, then rely on prompt instructions to avoid exposing sensitive values.
- D
Replace row access and masking policies with tags in the semantic model, because Cortex Analyst enforces tag-based restrictions instead of underlying Snowflake data policies.
Show answer and explanation
Correct answer: A
Explanation
To support secure data exploration with Cortex Analyst, organizations should keep enforcement in Snowflake's native security layer. If the underlying tables or views are protected by row access policies, masking policies, and appropriate grants, those controls continue to govern what a user can see when asking natural-language questions. Using governed views or directly governed tables in the semantic model is the practical pattern because it preserves centralized policy enforcement and least privilege. Security engineers should avoid creating ungoverned copies of data, elevating privileges for convenience, or assuming metadata tags or prompt instructions can replace enforced access controls. This approach is consistent with Snowflake best practices around policy-based access control, dynamic data masking, row-level security, and secure data access through governed objects.
- A. Correct.
Correct. The safest design is to expose Cortex Analyst to governed data sources that already enforce Snowflake security controls, such as row access policies, masking policies, and secure views where appropriate. When users access data through their own Snowflake identities and active roles, policy enforcement remains consistent with standard Snowflake authorization behavior. This aligns with the principle that AI-assisted exploration should not bypass existing data governance.
- B. Incorrect.
Incorrect. Exporting data into an unprotected schema removes the existing governance boundary and creates a high-risk copy of sensitive data. Cortex Analyst does not replace Snowflake's native security model with an independent protection layer. A common misconception is that an AI interface automatically re-applies equivalent controls, but secure access should continue to rely on Snowflake roles, policies, and governed objects.
- C. Incorrect.
Incorrect. Granting broad privileges contradicts least-privilege design and can cause users to see data they would not normally be entitled to access. Prompt instructions are not a security control and should never be treated as a substitute for enforced authorization. This option reflects a common but dangerous misunderstanding that application-layer guidance can replace database-level access enforcement.
- D. Incorrect.
Incorrect. Tags are useful for classification, governance workflows, and can drive certain policy management patterns, but they do not by themselves replace row access policies or masking policies for query-time enforcement in this scenario. Cortex Analyst relies on the underlying Snowflake access model; it does not natively convert semantic-model tags into row-level or column-level security enforcement.