5.1Standard
The system must limit harmful input and output.
Avoid one filter.
Choose two designs.
× Incorrect
Combine input checks, prompt constraints, retrieval authorization, output checks, and human review, measuring false positives, bypasses, language gaps, and impact.
Detailed explanation
CorrectCombine input, prompt, retrieval, output, and human-review controls
Risk is stopped at several stages.
CorrectMeasure false positives, bypasses, language differences, delay, and user impact
Guardrail side effects are managed.
IncorrectIgnore false positives and make controls stricter without review
Legitimate use is harmed.
IncorrectOne output filter blocks every attack
Risks span layers.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01 Domain 4・5.2のガードレール、入力出力検査、人的監督を確認する。
Official references
What to confirm
安全対策を一つの分類器に依存せず、段階ごとに評価・改善できる。
Key points
- Guardrail
- Defense in depth
- False positive
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
Guardrail
A control that restricts unsafe inputs, outputs, actions, or model behavior.
Defense in depth
Using multiple independent controls so one failure does not remove all protection.
Question IDAWS-AIF-460
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