HealthAI Diagnostics plans to publish a Model Card for a radiology model before promoting it from a SageMaker Model Registry staging package to production. The compliance team requires that the Model Card contains an intended use statement, evaluation results by demographic slices, and a risk rating with mitigation steps. Which practice will best satisfy the requirement and streamline operational use with SageMaker?

Attaching a Model Card JSON document directly to the SageMaker Model Package in the Model Registry is the best practice because it ensures the Model Card, containing all required information (intended use, per-slice evaluation metrics, risk ratings, and mitigations), travels with the model throughout its lifecycle. This streamlines operational use by making the critical compliance information immediately accessible and associated with the specific model version during promotion from staging to production. Storing a plain-text summary in S3 requires manual lookup and isn't integrated. Logging metrics to CloudWatch provides data but not a structured Model Card. Relying on Model Monitor for auto-generation after deployment is too late for pre-promotion compliance checks.