Skip to content

Data card template: make a dataset legible before review

A data card is a short, source-backed description of a collection before a sample review.

It tells a reader what the collection is, how it was assembled, what it covers, which formats and units it uses, what is known to be missing, and what rights or handling questions remain. It should make uncertainty visible. It should not turn a description into a quality guarantee or legal clearance.

  • The research basis for this structure is the datasheet practice of documenting motivation, composition, collection process, and recommended uses. That research is a documentation framework, not a legal standard and not a Rightmark certification. Gebru et al., (opens in a new tab)
  • The Rightmark Buyers page is the public source for the current data-card and sample-review boundary. This template is an editorial aid, not an additional product capability. Rightmark,

Updated

What is your data worth to AI developers?

Appraise your data
On this page

Template

Summary

  • Name: [neutral collection label]
  • Version or snapshot: [identifier or unknown]
  • Owner role: [role, not a private person’s name]
  • Last reviewed: [date or not reviewed]

Motivation and intended use

  • Why was the collection assembled? [source-backed description]
  • What uses are in scope? [description or needs review]
  • What uses are out of scope? [description or needs review]

Intended use is a review input. It does not grant permission to use the collection.

Composition and coverage

  • Source systems: [system classes]
  • Period covered: [period or unknown]
  • Languages: [languages or unknown]
  • Domains or subjects: [scope or unknown]
  • Formats and units: [description]
  • Labels or metadata: [description or none known]

Collection and processing

  • Collection process: [source-backed description]
  • Filtering, redaction, or transformation: [description]
  • Exclusions: [description]
  • Known gaps: [description]

Rights, privacy, and handling

  • Rights evidence reference: [internal ref or not supplied]
  • Notice or consent reference: [internal ref or not supplied]
  • Customer, staff, privileged, or regulated exclusions: [description]
  • Access and retention note: [description or needs review]

This section records evidence and open questions. It is not a legal conclusion. Health data remains research and status only and must not be collected through this template.

Rightmark does not take health data yet (patient records, medical imaging or veterinary records), until a HIPAA program and vendor agreements are in place.

General information, not legal advice.

Sample and review

  • Sample shape: [redacted schema or no sample]
  • Review method: [description]
  • Known limitations: [description]
  • Reviewer role: [role]

Change log and sources

  • Change: [what changed]
  • Date: [date]
  • Source references: [links or internal evidence refs]

Code collection note

For a code collection, keep the data card separate from the bounded Repo Check payload. The README shipped with the Repo Check download describes a data card with fixed words and numbers, estimated tokens, history size and span, languages and line counts, engineering signals, provenance markers, dependency families, privacy counts, and known gaps. It also states that the data card holds no names, paths, email addresses, commit messages, code, or dollar figure. Follow the current CLI output rather than copying fields into a hand-built result.

Safe example

  • Dataset A
  • Purpose: [source-backed description]
  • Coverage: [period and scope]
  • Format: [format and unit]
  • Process: [source-backed description]
  • Known gaps: [explicit gaps]
  • Rights evidence: [internal reference]
  • Privacy handling: [exclusions and access note]
  • Review status: Needs evidence

The example is a blank synthetic shape. It is not an existing dataset, customer proof, quality score, or availability statement.

Sources

  1. Gebru et al.: Datasheets for Datasets, Communications of the ACM (opens in a new tab)
  2. Rightmark: Buyers
  3. Rightmark: Repo Check

Use the Buyers page for the current data-card and sample-review boundary.

Reports, such as Repo Check numbers or a system inventory, and representative scrubbed samples under NDA, for evaluation only, so your team can test fit first.