The correct option is Output explainability. This capability provides per-result rationales that help auditors understand why specific clauses were omitted from a generated summary.

The feature returns human readable explanations tied to individual outputs so reviewers can see which input segments, tokens, or reasoning steps influenced the model decision. That level of per-output rationale is what auditors need to trace why a particular clause was not included in a condensed policy summary.

Amazon SageMaker Model Cards document model metadata and governance artifacts and they support lifecycle transparency but they do not produce justifications for individual generated outputs.

Amazon Bedrock Guardrails enforce safety and compliance constraints and they help shape or limit outputs but they do not provide the per-result rationale that explains why a clause was excluded.

Model fine-tuning can bias a model toward desired behavior and improve content inclusion patterns but it does not inherently produce human readable explanations tied to each specific summary.

Full AWS Practitioner Certification Question