Question 10
Domain 5: GovernanceA team is deploying a Gen AI application and wants to reduce harm from malicious user prompts before the model processes them. Which control is the best example of a guardrail technique for this purpose?
Correct answer: A
Explanation
Guardrail techniques are controls applied to Gen AI inputs or interactions to detect, block, or constrain malicious user content before it can cause harmful model behavior. — official.txt
Why each option is right or wrong
A. Screen incoming prompts and block or sanitize malicious content before submission to the model.
The topic specifically identifies guardrail techniques as protections against malicious user inputs to a Gen AI application, so filtering or sanitizing prompts before model processing directly matches that purpose.
B. Increase the model’s parameter count so it can better tolerate adversarial or abusive prompts.
Guardrail techniques protect against malicious inputs; model size is not identified as an input protection control.
C. Reduce network latency so user requests reach the application more quickly during peak periods.
Performance optimization addresses speed, not protection against malicious user inputs.
D. Expand the application’s context window so longer prompts can be accepted without truncation.
Accepting longer prompts changes capacity, not detection or prevention of malicious user content.