Data Card
A data card is a structured document describing a dataset used to train or evaluate an AI model: its source, composition, collection process, intended use, and limitations.
What is Data Card?
Data cards extend the model card idea to datasets. They typically describe collection methodology, demographic composition, labelling process, known biases, licensing, and recommended evaluation slices. Data cards are essential for downstream users to assess fitness for purpose. The EU AI Act explicitly requires data governance documentation for high-risk AI.
How does Data Card apply to enterprise AI?
Enterprises that fine-tune on internal data should produce data cards for those datasets, both for compliance and for future maintainers who will need to know what is in them.
Related terms
- Model Card - A model card is a structured document describing an AI model's purpose, training data, performance, limitations, and ethical considerations.
- AI Risk Management - AI risk management is the discipline of identifying, assessing, mitigating, and monitoring the harms an AI system can cause across its lifecycle.
- Data Residency - Data residency is the requirement that personal or regulated data stays within a specified geographic region throughout processing, storage, and backup.
- EU AI Act - The EU AI Act (Regulation (EU) 2024/1689) is the European Union's horizontal regulation for AI, classifying systems by risk and imposing obligations on providers, deployers, importers, and distributors.
External references
Need help applying Data Card to your enterprise? Submit a short brief and we reply within one business day.