An AI repeatedly returns dangerous advice and impact is growing.
Stop harm while investigating.
Which initial action is most appropriate?
Stop the feature or traffic, preserve evidence, and move to notice, investigation, and recovery decisions.
Detailed explanation
Impact is contained while investigation continues.
Impact is contained while investigation continues.
Harm can grow.
Harm can grow.
Investigation and accountability are damaged.
Investigation and accountability are damaged.
Users are not adequately protected.
Users are not adequately protected.
Try it yourself
An example you can run in a temporary verification environment.
AWS公式AIF-C01 Domain 4.3の停止、インシデント対応、運用ガバナンスを確認する。Expected result
原因調査を待たずに影響を抑え、証跡と復旧判断を維持できる。Key points
- Kill switch
- Containment
- Evidence preservation
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
- Command output formatting can vary slightly by distribution or tool version.
- Run the example in a temporary directory or process when possible.
Foundation review
Read the scope first
Check whether the command acts on the current shell, a new process, an existing process, or a file.
Verify the observable result
Use the supplied command and compare the output with the expected result.