Users enter free-form text into an AI application.
Reject excessive, prohibited, or malicious content before generation.
Choose two appropriate controls.
Validate length, format, data classification, and harmful content, then route rejected input to safe handling.
Detailed explanation
The policy is bypassed and the failure is hidden.
Unexpected input and cost or leakage risk are reduced.
The application can branch before the model sees the input.
Injection and privilege escalation are possible.
Check official documentation
Compare the following points with the official documentation.
AWS公式生成AIセキュリティ・GuardrailsとAIF-C01 Domain 2.3を確認する。
Official references
What to confirm
入力検証をコスト、安全性、権限の境界として設計できる。
Key points
- Input validation
- Data class
- Reject
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
Input validation
Checking data type, length, format, content, sensitivity, and scope before processing.
Input policy
Rules describing which data, purposes, users, and contents may be sent to an AI system.