Model card

A standardized documentation artifact describing a trained model's intended use, training data at a summary level, performance characteristics, and known limitations. A model card that honestly documents a model's training data sources and any known retention or lawful-basis gaps is a natural home for the retention feasibility findings this topic's lab produces, complementing the deeper training data summary treatment at (see Topic 8.2).

Structured documentation released alongside a machine learning model that reports its intended use, known limitations, and stratified performance data, a practice established by Margaret Mitchell and colleagues at Google in 2019 and now common among major model providers as a way to make bias-relevant evidence available before deployment rather than only discoverable after.

A short, structured document that travels with a trained model and states, in plain language, what the model is, what it is for, where it works, where it fails, and how that is known. Introduced formally by Mitchell and colleagues (2019). Covers model details, intended and out-of-scope use, factors, metrics, evaluation and training data, disaggregated results, and caveats.

A short, standardized document published alongside an AI model that describes its intended use, known limitations, training data at a high level, and evaluation results. Model cards are a common developer-side transparency mechanism referenced in several current and proposed AI disclosure laws.

A documentation format (popularized by Google, 2019) giving structured information about a model's intended use, performance, training data, limitations, and ethical considerations; documentation of this kind is required for high-risk systems under the EU AI Act.