3.4Standard
Test a chatbot's safety.
Use normal and adversarial inputs.
Choose two evaluations.
× Incorrect
Measure harmful categories, false positives, false negatives, paraphrases, prompts, and language differences rather than testing only normal questions.
Detailed explanation
IncorrectTest safety only with ordinary questions
Adversarial input is missing.
IncorrectOne filter pass proves safety
Behavior changes with context and phrasing.
CorrectMeasure harmful, discriminatory, sexual, self-harm, and privacy-request categories
Risk-specific weaknesses are visible.
CorrectTest false refusals, paraphrases, coercion, and language differences
Both overblocking and misses are measured.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01 Domain 3.4の安全性評価、コンテンツフィルター、レッドチームを確認する。
Official references
What to confirm
有害出力と過剰拒否を複数ケースで評価できる。
Key points
- Harm
- False refusal
- Red team
Notes
- Environment: AWS公式AIF-C01試験ガイドとAWS公式ドキュメントの確認
- Compare the definition with the AIF-C01 objective and the service documentation before applying it.
- Keep the concept separate from a particular answer choice or implementation step.
Foundation review
False refusal
Rejecting a safe request that policy allows the system to answer.
Red-team evaluation
Deliberately probing a system with adversarial or boundary inputs to find weaknesses.
Question IDAWS-AIF-343
EnvironmentAWS公式AIF-C01試験ガイドとAWS公式ドキュメントの確認
Last technical review2026-08-20
Report an error or rights issue