Answers should hide customer emails. Which technique fits?
Why D is the answer
Masking
Masking replaces sensitive values with placeholders, so an answer stays useful with no private data.
In detail
- Masking is a guardrail technique.
- Masking hides values while the rest of the answer stays useful.
- Mask before logging, as well as before display.
Use case
An illustrative scenario, not a real incident.
A support summary should not show customer emails. Masking replaces them with a placeholder, so the summary stays useful and the address stays private.
See it in different ways
flowchart LR A["User input"] --> B["Input guardrail"] B --> C["LLM"] C --> D["Mask or filter output"] D --> E["User"]
mindmap
root((Governance))
Input
Guardrails
Output
Masking
Data
Licensing
Permissionsflowchart LR S0["Generated text"] --> S1["Detect emails"] --> S2["Mask"] --> S3["Show to the user"]
Key point: Masking hides values and keeps the text useful.
Why the other options are wrong
- A. Larger chunks
- Chunk size does not hide values.
- B. Higher temperature
- Temperature does not remove sensitive data.
- C. More reranking
- Reranking changes order, and the data remains.

