In plain English
A model card is a standardised document that accompanies an AI model and describes its intended use, performance, limitations, training data, and ethical considerations. It is meant to travel with the model so the people deploying it can judge whether it fits their context. Think of it as a nutrition label for a model.
Why it matters
A model card helps a deploying organisation make an informed choice and document due diligence, rather than relying on marketing claims about what a model can do.
A worked example
A model card might note that an image classifier was trained mainly on daytime photos and performs poorly at night, warning against using it for nighttime footage.
Common confusion
A model card describes the model in general; it is not a guarantee of safe performance in your specific deployment, which still needs its own assessment.