Practice
Learn
← Quit
0 / 0 correct
LLMs & AI
Hard
1 of 20
How should guardrails be layered around an LLM application?
A
Only filter outputs with regex
B
Only add a polite system prompt
C
Validate and filter inputs, constrain tools and data access, check outputs (schema, safety, PII) before acting or displaying, and log for audit
D
Trust the model's refusals