Module 12 Knowledge Check
5 questions. Pass mark 4 out of 5. Answer every question before checking the answer key below, then retry after reading the feedback.
1. Why do retrieval tables for AI need explicit permission columns?
- A. To save space
- B. So an assistant cannot surface records a given user is not allowed to see
- C. To speed up joins
- D. They do not
2. Useful audit fields on a table feeding an AI system include…
- A. Only a row id
- B. Source, created/updated timestamps, and a version or checksum
- C. Only the text
- D. Random ids
3. A retrieval row reads only Status completed. Total 1000.00. Why is that a defect?
- A. It is too short to embed
- B. Retrieved on its own it does not name its subject, so the model cannot tell which order it describes
- C. It should be stored as JSON
- D. Decimal values are not allowed in retrieval content
4. Stale rows in a retrieval table are dangerous because…
- A. They use storage
- B. The assistant will cite outdated information as current fact
- C. They slow queries
- D. They cannot be deleted
5. Before exposing a table to an AI workflow you should confirm…
- A. It has an index
- B. That it contains no personal or confidential fields the use case does not require
- C. That it is sorted
- D. That it has under 1M rows
---
Answer Key and Explanations
Check these only after attempting every question.
1. B - So an assistant cannot surface records a given user is not allowed to see
Retrieval systems will happily return whatever they are given; access control must be in the data layer.
2. B - Source, created/updated timestamps, and a version or checksum
Without provenance you cannot explain, reproduce, or retract an answer the assistant gave.
3. B - Retrieved on its own it does not name its subject, so the model cannot tell which order it describes
Every retrievable row must name its own subject and carry a pointer back to the source record. Order 501 placed on 2026-06-21 for IN. Status completed. Total 1000.00. survives retrieval alone; the fragment does not, and the model will attach it to the wrong order.
4. B - The assistant will cite outdated information as current fact
Freshness is a correctness property once data is being quoted back to users.
5. B - That it contains no personal or confidential fields the use case does not require
Data minimisation limits the blast radius if the assistant leaks or mis-scopes a result.
Practical Check
Apply this module to your own work: complete the module activity for *AI Data Preparation: Retrieval Tables, Metadata, Permissions, and Audit Fields*, then write one sentence naming what your result shows and one naming what it does not.
Strong Answer Pattern
A strong answer names the task, the evidence used, the check performed, and the remaining limitation. It avoids "proved", "guaranteed", or "always" unless the evidence genuinely supports it.