An AI application needs an audit trail of supported AWS API activity and, where permitted, records of model inputs and outputs for quality investigation. Which logging considerations are appropriate? Select 2 answers.

1. Use CloudTrail for supported API activity and separately evaluate Bedrock model invocation logging for supported interaction payloads.
2. Apply access, encryption, retention, and data-minimization controls to logs because prompts and responses can contain sensitive information.
3. Make all prompt logs publicly readable so reviewers do not need authorization.
4. Retain every interaction forever without considering user privacy or retention policies.
5. Assume every CloudTrail event always includes the complete prompt and generated answer for every model API.