COMPEL Glossary / model-card
Model Card
A model card is a standardized documentation template that describes an AI model's intended use, training data, performance characteristics across different populations, known limitations, fairness evaluations, and ethical considerations.
What this means in practice
Model cards support transparency by providing stakeholders with the information needed to assess whether an AI system is appropriate for a given use case. For regulators, model cards demonstrate governance rigor. For internal teams, they enable informed decisions about model deployment and usage. Model cards should be updated whenever a model is retrained or its deployment context changes. In the COMPEL framework, model cards are part of the Deployed System Record (TMPL-P-001) and must be maintained as living documents throughout the model's production lifecycle.
Why it matters
Model cards support transparency by providing stakeholders with standardized information needed to assess whether an AI system is appropriate for a given use case. For regulators, model cards demonstrate governance rigor. For internal teams, they enable informed deployment decisions. Without model cards, organizations cannot efficiently communicate model characteristics, creating knowledge gaps that lead to inappropriate usage and governance failures.
How COMPEL uses it
Model cards are part of the Deployed System Record (TMPL-P-001), a mandatory COMPEL governance artifact maintained throughout the model's production lifecycle. During Model, model card templates are established as part of governance standards. The Produce stage creates model cards for each deployed model. The Evaluate stage audits card completeness and accuracy. Cards must be updated whenever models are retrained or deployment contexts change.
Related Terms
Other glossary terms mentioned in this entry's definition and context.