Datasheet for Datasets

Full definition

Datasheets sit beneath model cards in the documentation stack: a model card describes the trained artifact, while a datasheet describes the data it learned from. Critical fields include: who collected the data, what consents were obtained, demographic distributions, label sources and quality, and known biases. The EU AI Act's Article 10 effectively requires datasheet-equivalent disclosures for high-risk systems.

Why it matters

Most AI failures trace to data, not algorithms. A datasheet exposes the exact failure modes — coverage gaps, consent ambiguity, label noise — that downstream model cards inherit. Buyers should refuse training-data summaries and demand datasheets.

Example

A facial-recognition vendor publishes a datasheet revealing 86% of training images came from one continent; a procurement committee disqualifies the product on coverage grounds before pilot.

Related terms

• Model Card
  A model card is a short, structured document that describes an AI model's intended use, performance, limitations, training data, evaluation results across demographic and operational subgroups, and known failure modes — the AI equivalent of a nutrition label.
• Training Data
  Training data is the curated dataset used to teach a machine-learning model — every pattern the model can recognize, every bias it inherits, and every limit to its accuracy ultimately traces back to this data.
• AI Bias
  AI bias is systematic, unfair difference in an AI system's outputs across demographic, geographic, or other groups — usually caused by biased training data, biased labels, or biased problem framing rather than the algorithm itself.
• AI Governance
  AI governance is the framework of policies, roles, controls, and processes an organization uses to ensure its AI systems are lawful, safe, fair, accountable, and aligned with business intent — across the full lifecycle from problem framing to retirement.