Question 360
At fictional Willowhaven Municipal Office, a Snowflake data and AI engineering team is conducting a source-system modernization and AI search review while preserving enterprise data boundaries and a reviewable audit trail. The separately composed technical incident is: A text-to-SQL assistant exposes descriptions of restricted financial tables even when the requesting analyst cannot query those tables, leaking sensitive schema context. Assume only these documented facts and use safe, authorized production-change procedures; exact current Snowflake feature syntax and region or account availability require current review. Which repeatable assurance or control is most relevant to preventing this situation?
Explanation
Writer-proposed answer: D (UNVERIFIED)
Preliminary individual option explanations — NOT expert validated
Correct Answer
Audit semantic metadata grants together with source-table permissions and run role-specific discovery tests after updates. Metadata can itself be sensitive; text-to-SQL systems must enforce access at both semantic-context and query execution stages.
Incorrect Answers
This response concerns Unreviewed language model version drift, which is different from the stated failure. A text-to-SQL assistant exposes descriptions of restricted financial tables even when the requesting analyst cannot query those tables, leaking sensitive schema context.
LLM observability without protected prompt leakage is insufficient to demonstrate this result: Limited-role tests reveal neither restricted table descriptions nor data and reject generated queries that would cross the caller's authorization boundary.
Cortex model access allowlist governance is not the cause of the reported failure: A text-to-SQL assistant exposes descriptions of restricted financial tables even when the requesting analyst cannot query those tables, leaking sensitive schema context.