A company deploying a generative AI assistant wants to address responsible-AI concerns. Which TWO practices directly reduce the risk of harmful or biased outputs? (Select TWO.)
Choose 2 answers.
A.Apply Bedrock Guardrails for content filtering and denied topics
Correct. Guardrails are the managed safety layer.
B.Evaluate model outputs for bias across demographic groups before launch
Correct. Fairness evaluation is core to responsible AI.
C.Increase the model's temperature
Higher randomness increases unpredictability, not safety.
D.Deploy without human review to reduce latency
Human-in-the-loop review is a key mitigation — removing it raises risk.
E.Hide from users that they're interacting with AI
Transparency/disclosure is a responsible-AI requirement, not a hazard to remove.
Why this answer
Guardrails enforce safety policies at inference (content filters, denied topics, PII redaction); bias evaluation measures fairness across groups. Higher temperature, removing human oversight, and concealing AI use all increase risk — not reduce it.
Correct answers: A, B · Last reviewed 2026-10-26
Official AWS references
Practice this interactively
Answer without seeing the solution first in the Responsible AI practice session.
Practice Responsible AI