Last updated:
A short, standardized document providing essential information about a machine learning model's performance, limitations, and intended use cases.
Think of a Model Card as a nutrition label for AI. It typically includes accuracy benchmarks across different demographic groups, descriptions of the training data, and explicit warnings about out-of-scope use cases. For third-party AI, it is the primary evidence a deployer uses to satisfy transparency requirements under the EU AI Act.
Real world example:
A vendor providing a customer sentiment analyzer gives the buyer a Model Card showing the system is 95% accurate for English but only 60% for regional dialects - prompting the buyer to implement extra human review for those regions.
