A public AI application sees long inputs and attempts to evade refusals.
Detect abuse while preserving legitimate usage.
Choose two appropriate controls.
Monitor per-user and per-feature rate, input length, refusals, and cost, then apply quotas, limits, blocks, and human review with protected logs.
Detailed explanation
Abuse is reduced without treating every user identically.
Abnormal patterns and cost spikes become visible.
Monitoring data also needs minimization and protection.
Cost, leakage, and unsafe actions become more likely.
Check official documentation
Compare the following points with the official documentation.
AWS公式AIF-C01 Domain 3.4・5.1の監視、レート制限、AIセキュリティを確認する。
Official references
What to confirm
濫用指標と段階的な制御、ログ最小化を説明できる。
Key points
- Abuse
- Quota
- Log protection
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
Prompt abuse
Excessive, evasive, or unauthorized use of prompts, generation, or connected tools.
Quota
A limit on requests, tokens, concurrency, storage, or spend for an account or principal.