An ML team shares model artifacts with other teams but retains training code and data. They want a publish-time mechanism to support auditing and transparency for their custom models. Which solution should they use when publishing those models?

Amazon SageMaker Model Cards provide a standardized way to document model details, including intended uses, training data, and inference information. This directly addresses the need for a publish-time mechanism to support auditing and transparency for custom models, even when the training code and data are not shared. Writing documentation to S3 is a manual process and lacks the structured, standardized format of Model Cards. AWS AI Service Cards are for AWS pre-built AI services, not custom models. Providing training scripts in a Git repository doesn't offer the structured transparency and auditing information that Model Cards do for the published model artifact itself.