What each data card standard asks a dataset to document

Name: What each data card standard asks a dataset to document
Creator: Data Card Generator
Published: 2026-10-08
License: https://datacardgenerator.com/terms

6 published standards and legal templates say what a dataset's documentation must contain, each quoted from its own text, read 8 October 2026.

6 standards Standards read field by field 11 texts checked, 8 Oct 2026

3 of 6 Binding or voluntary: a voluntary standard or format the answer given most, of 3

4 of 6 Asks where the data came from: asks for the data's sources or provenance the answer given most, of 2

4 of 6 Asks about personal data: not stated the answer given most, of 2

6 of 6 What it asks the card to contain: stated how many standards say so in their texts, quoted below

11 Texts examined 15 named sources; every answer is dated and hashed

Binding or voluntary

a voluntary standard or format3 standards
a legal obligation2 standards
not stated1 standard

Across 6 standards, a single answer each.

Counts

Row	Value
a voluntary standard or format	3 standards
a legal obligation	2 standards
not stated	1 standard

Asks where the data came from

asks for the data's sources or provenance4 standards
not stated2 standards

Across 6 standards, a single answer each.

Counts

Row	Value
asks for the data's sources or provenance	4 standards
not stated	2 standards

Asks about personal data

not stated4 standards
asks whether personal data is included2 standards

Across 6 standards, a single answer each.

Counts

Row	Value
not stated	4 standards
asks whether personal data is included	2 standards

What it asks the card to contain

stated6 standards

Across 6 standards, a single answer each.

Counts

Row	Value
stated	6 standards

The full board, one standard per row

One standard to a row. Expand an answer for its source sentence, word for word, plus the text and the read date; "not stated" means that text does not cover the question.

Standard	Binding or voluntary	Asks where the data came from	Asks about personal data	What it asks the card to contain
California AB 2013	a legal obligation On or before January 1, 2026, and before each time thereafter that a generative artificial intelligence system or service California AB 2013 bill text	asks for the data's sources or provenance (1) The sources or owners of the datasets. California AB 2013 bill text	asks whether personal data is included (7) Whether the datasets include personal information, as defined in subdivision (v) of Section 1798.140. California AB 2013 bill text	A high-level summary of the datasets used in the development of the generative artificial… A high-level summary of the datasets used in the development of the generative artificial intelligence system or service, including, but not limited to: California AB 2013 bill text
Croissant (MLCommons)	a voluntary standard or format Not addressed in Croissant format specification Croissant is an open, community-built, standardized metadata vocabulary for ML datasets MLCommons Croissant working group page Not addressed in Google Research blog: Croissant	asks for the data's sources or provenance Not addressed in Croissant format specification Croissant provides machine-readable descriptions of dataset provenance and permissions MLCommons Croissant working group page Not addressed in Google Research blog: Croissant	not stated Not addressed in Croissant format specification Not addressed in MLCommons Croissant working group page Not addressed in Google Research blog: Croissant	It provides a vocabulary for dataset attributes, streamlining how data is loaded across… It provides a vocabulary for dataset attributes, streamlining how data is loaded across ML frameworks such as PyTorch, TensorFlow or JAX. Croissant format specification including key attributes and properties of datasets, as well as information required to load them into ML tools MLCommons Croissant working group page This 1.0 release of Croissant includes a complete specification of the format Google Research blog: Croissant
Data Cards (Google Research)	not stated Not addressed in Data Cards paper (arXiv 2204.01075) Not addressed in Data Cards Playbook	asks for the data's sources or provenance A clear and thorough understanding of a dataset's origins, development, intent, ethical considerations and evolution Data Cards paper (arXiv 2204.01075) Is this source considered sensitive or high-risk? [Yes/No] Data Cards Playbook	asks whether personal data is included Not addressed in Data Cards paper (arXiv 2204.01075) (S/PII were collected as a part of the dataset creation process.) Data Cards Playbook	A clear and thorough understanding of a dataset's origins, development, intent, ethical… A clear and thorough understanding of a dataset's origins, development, intent, ethical considerations and evolution Data Cards paper (arXiv 2204.01075) Data fields that were collected and are included in the dataset. Data Cards Playbook
Datasheets for Datasets	a voluntary standard or format we propose datasheets for datasets Datasheets for Datasets paper (arXiv 1803.09010)	asks for the data's sources or provenance we propose that every dataset be accompanied with a datasheet that documents its motivation, composition, collection process, recommended uses Datasheets for Datasets paper (arXiv 1803.09010)	not stated Not addressed in Datasheets for Datasets paper (arXiv 1803.09010)	we propose that every dataset be accompanied with a datasheet that documents its… we propose that every dataset be accompanied with a datasheet that documents its motivation, composition, collection process, recommended uses Datasheets for Datasets paper (arXiv 1803.09010)
EU AI Act training-content summary (Art. 53(1)(d) and template)	a legal obligation Not addressed in Commission: explanatory notice and template Providers of general-purpose AI models shall: AI Act Article 53 (European Commission AI Act Service Desk)	not stated Not addressed in Commission: explanatory notice and template Not addressed in AI Act Article 53 (European Commission AI Act Service Desk)	not stated Not addressed in Commission: explanatory notice and template Not addressed in AI Act Article 53 (European Commission AI Act Service Desk)	The Template annexed to this Explanatory Notice aims to provide a common minimal baseline… The Template annexed to this Explanatory Notice aims to provide a common minimal baseline for the information to be made publicly available in the Summary of Training Content Commission: explanatory notice and template draw up and make publicly available a sufficiently detailed summary about the content used for training of the general-purpose AI model, according to a template provided by the AI Office. AI Act Article 53 (European Commission AI Act Service Desk)
Hugging Face dataset card	a voluntary standard or format To control how the Hub displays the card, you should create a YAML section in the README file to define some metadata. Hugging Face Hub docs: dataset cards Not addressed in hub-docs datasetcard.md (GitHub)	not stated Not addressed in Hugging Face Hub docs: dataset cards Not addressed in hub-docs datasetcard.md (GitHub)	not stated Not addressed in Hugging Face Hub docs: dataset cards Not addressed in hub-docs datasetcard.md (GitHub)	A dataset repo will render its README.md as a dataset card. A dataset repo will render its README.md as a dataset card. Hugging Face Hub docs: dataset cards For a template for the human-readable portion of the dataset card, see: datasetcard_template.md file. hub-docs datasetcard.md (GitHub)

Quoted wording for all 44 answers

California AB 2013, California AB 2013 bill text, read 2026-10-08, Binding or voluntary: On or before January 1, 2026, and before each time thereafter that a generative artificial intelligence system or service
California AB 2013, California AB 2013 bill text, read 2026-10-08, Asks where the data came from: (1) The sources or owners of the datasets.
California AB 2013, California AB 2013 bill text, read 2026-10-08, Asks about personal data: (7) Whether the datasets include personal information, as defined in subdivision (v) of Section 1798.140.
California AB 2013, California AB 2013 bill text, read 2026-10-08, What it asks the card to contain: A high-level summary of the datasets used in the development of the generative artificial intelligence system or service, including, but not limited to:
Croissant (MLCommons), Croissant format specification, read 2026-10-08, Binding or voluntary: Nothing on this text answers it (64,269 characters of text text, read 2026-10-08).
Croissant (MLCommons), Croissant format specification, read 2026-10-08, Asks where the data came from: Nothing on this text answers it (64,269 characters of text text, read 2026-10-08).
Croissant (MLCommons), Croissant format specification, read 2026-10-08, Asks about personal data: Nothing on this text answers it (64,269 characters of text text, read 2026-10-08).
Croissant (MLCommons), Croissant format specification, read 2026-10-08, What it asks the card to contain: It provides a vocabulary for dataset attributes, streamlining how data is loaded across ML frameworks such as PyTorch, TensorFlow or JAX.
Croissant (MLCommons), MLCommons Croissant working group page, read 2026-10-08, Binding or voluntary: Croissant is an open, community-built, standardized metadata vocabulary for ML datasets
Croissant (MLCommons), MLCommons Croissant working group page, read 2026-10-08, Asks where the data came from: Croissant provides machine-readable descriptions of dataset provenance and permissions
Croissant (MLCommons), MLCommons Croissant working group page, read 2026-10-08, Asks about personal data: Nothing on this text answers it (10,613 characters of text text, read 2026-10-08).
Croissant (MLCommons), MLCommons Croissant working group page, read 2026-10-08, What it asks the card to contain: including key attributes and properties of datasets, as well as information required to load them into ML tools
Croissant (MLCommons), Google Research blog: Croissant, read 2026-10-08, Binding or voluntary: Nothing on this text answers it (8,898 characters of text text, read 2026-10-08).
Croissant (MLCommons), Google Research blog: Croissant, read 2026-10-08, Asks where the data came from: Nothing on this text answers it (8,898 characters of text text, read 2026-10-08).
Croissant (MLCommons), Google Research blog: Croissant, read 2026-10-08, Asks about personal data: Nothing on this text answers it (8,898 characters of text text, read 2026-10-08).
Croissant (MLCommons), Google Research blog: Croissant, read 2026-10-08, What it asks the card to contain: This 1.0 release of Croissant includes a complete specification of the format
Data Cards (Google Research), Data Cards paper (arXiv 2204.01075), read 2026-10-08, Binding or voluntary: Nothing on this text answers it (3,651 characters of text text, read 2026-10-08).
Data Cards (Google Research), Data Cards paper (arXiv 2204.01075), read 2026-10-08, Asks where the data came from: A clear and thorough understanding of a dataset's origins, development, intent, ethical considerations and evolution
Data Cards (Google Research), Data Cards paper (arXiv 2204.01075), read 2026-10-08, Asks about personal data: Nothing on this text answers it (3,651 characters of text text, read 2026-10-08).
Data Cards (Google Research), Data Cards paper (arXiv 2204.01075), read 2026-10-08, What it asks the card to contain: A clear and thorough understanding of a dataset's origins, development, intent, ethical considerations and evolution
Data Cards (Google Research), Data Cards Playbook, read 2026-10-08, Binding or voluntary: Nothing on this text answers it (46,306 characters of text text, read 2026-10-08).
Data Cards (Google Research), Data Cards Playbook, read 2026-10-08, Asks where the data came from: Is this source considered sensitive or high-risk? [Yes/No]
Data Cards (Google Research), Data Cards Playbook, read 2026-10-08, Asks about personal data: (S/PII were collected as a part of the dataset creation process.)
Data Cards (Google Research), Data Cards Playbook, read 2026-10-08, What it asks the card to contain: Data fields that were collected and are included in the dataset.
Datasheets for Datasets, Datasheets for Datasets paper (arXiv 1803.09010), read 2026-10-08, Binding or voluntary: we propose datasheets for datasets
Datasheets for Datasets, Datasheets for Datasets paper (arXiv 1803.09010), read 2026-10-08, Asks where the data came from: we propose that every dataset be accompanied with a datasheet that documents its motivation, composition, collection process, recommended uses
Datasheets for Datasets, Datasheets for Datasets paper (arXiv 1803.09010), read 2026-10-08, Asks about personal data: Nothing on this text answers it (2,911 characters of text text, read 2026-10-08).
Datasheets for Datasets, Datasheets for Datasets paper (arXiv 1803.09010), read 2026-10-08, What it asks the card to contain: we propose that every dataset be accompanied with a datasheet that documents its motivation, composition, collection process, recommended uses
EU AI Act training-content summary (Art. 53(1)(d) and template), AI Act Article 53 (European Commission AI Act Service Desk), read 2026-10-08, Binding or voluntary: Providers of general-purpose AI models shall:
EU AI Act training-content summary (Art. 53(1)(d) and template), AI Act Article 53 (European Commission AI Act Service Desk), read 2026-10-08, Asks where the data came from: Nothing on this text answers it (6,402 characters of text text, read 2026-10-08).
EU AI Act training-content summary (Art. 53(1)(d) and template), AI Act Article 53 (European Commission AI Act Service Desk), read 2026-10-08, Asks about personal data: Nothing on this text answers it (6,402 characters of text text, read 2026-10-08).
EU AI Act training-content summary (Art. 53(1)(d) and template), AI Act Article 53 (European Commission AI Act Service Desk), read 2026-10-08, What it asks the card to contain: draw up and make publicly available a sufficiently detailed summary about the content used for training of the general-purpose AI model, according to a template provided by the AI Office.
EU AI Act training-content summary (Art. 53(1)(d) and template), Commission: explanatory notice and template, read 2026-10-08, Binding or voluntary: Nothing on this text answers it (5,138 characters of text text, read 2026-10-08).
EU AI Act training-content summary (Art. 53(1)(d) and template), Commission: explanatory notice and template, read 2026-10-08, Asks where the data came from: Nothing on this text answers it (5,138 characters of text text, read 2026-10-08).
EU AI Act training-content summary (Art. 53(1)(d) and template), Commission: explanatory notice and template, read 2026-10-08, Asks about personal data: Nothing on this text answers it (5,138 characters of text text, read 2026-10-08).
EU AI Act training-content summary (Art. 53(1)(d) and template), Commission: explanatory notice and template, read 2026-10-08, What it asks the card to contain: The Template annexed to this Explanatory Notice aims to provide a common minimal baseline for the information to be made publicly available in the Summary of Training Content
Hugging Face dataset card, hub-docs datasetcard.md (GitHub), read 2026-10-08, Binding or voluntary: Nothing on this text answers it (2,066 characters of text text, read 2026-10-08).
Hugging Face dataset card, hub-docs datasetcard.md (GitHub), read 2026-10-08, Asks where the data came from: Nothing on this text answers it (2,066 characters of text text, read 2026-10-08).
Hugging Face dataset card, hub-docs datasetcard.md (GitHub), read 2026-10-08, Asks about personal data: Nothing on this text answers it (2,066 characters of text text, read 2026-10-08).
Hugging Face dataset card, hub-docs datasetcard.md (GitHub), read 2026-10-08, What it asks the card to contain: For a template for the human-readable portion of the dataset card, see: datasetcard_template.md file.
Hugging Face dataset card, Hugging Face Hub docs: dataset cards, read 2026-10-08, Binding or voluntary: To control how the Hub displays the card, you should create a YAML section in the README file to define some metadata.
Hugging Face dataset card, Hugging Face Hub docs: dataset cards, read 2026-10-08, Asks where the data came from: Nothing on this text answers it (7,982 characters of text text, read 2026-10-08).
Hugging Face dataset card, Hugging Face Hub docs: dataset cards, read 2026-10-08, Asks about personal data: Nothing on this text answers it (7,982 characters of text text, read 2026-10-08).
Hugging Face dataset card, Hugging Face Hub docs: dataset cards, read 2026-10-08, What it asks the card to contain: A dataset repo will render its README.md as a dataset card.

Read and excluded

We read 4 named documents that turned out not to be what this board asks about; why is given for each below. They count toward the reading, not the answers.

European Union (official text): EUR-Lex answered a 'temporarily not fully available' notice; the Regulation's text is not in the stored read
NIST AI RMF: the NIST AI RMF landing page; it sets no dataset documentation fields in the stored read
[1810.03993] Model Cards for Model Reporting: Model Cards document trained models, not datasets
AI Act article text mirror: the charter admits it for: reading aid only; EUR-Lex is the authority

How this was built

Each standard's own official texts, templates and specifications were read through estate-fetch and stored whole (sha256 on every row).

For every text, four questions were answered from its own words: binding or voluntary, asks where the data came from, asks about personal data, what it asks the card to contain.

Every stated answer carries the sentence it rests on, checked against the stored text; 'not stated' means the text read does not address the question.

A standard with several texts reads 'varies' where they disagree.

6 standards, 11 texts.

All 15 sources

European Union (official text) (cited for reporting; the publisher's own official texts, templates and specifications)
Croissant Format Specification | Croissant site (cited for reporting; the publisher's own official texts, templates and specifications)
European Union (official text) (read, set aside)
[2204.01075] Data Cards: Purposeful and Transparent Dataset Documentation for Responsible (cited for reporting; the publisher's own official texts, templates and specifications)
The Data Cards Playbook - Data Cards Playbook (cited for reporting; the publisher's own official texts, templates and specifications)
NIST AI RMF (read, set aside)
Croissant - MLCommons (cited for reporting; the publisher's own official texts, templates and specifications)
Dataset Cards · Hugging Face (cited for reporting; the publisher's own official texts, templates and specifications)
California AB 2013 (cited for reporting; the publisher's own official texts, templates and specifications)
[1810.03993] Model Cards for Model Reporting (read, set aside)
AI Act article text mirror (read, set aside)
hub-docs/datasetcard.md at main · huggingface/hub-docs · GitHub (cited for reporting; the publisher's own official texts, templates and specifications)
[1803.09010] Datasheets for Datasets (cited for reporting; the publisher's own official texts, templates and specifications)
Croissant: a metadata format for ML-ready datasets (cited for reporting; the publisher's own official texts, templates and specifications)
European Commission AI Act Service Desk, Article 53: obligations for providers of general-purpose AI models (cited for reporting; the publisher's own official texts, templates and specifications)

Reference or embed these figures

6 published standards and legal templates say what a dataset's documentation must contain, each quoted from its own text, read 8 October 2026.

Binding or voluntary, by standard: a voluntary standard or format 3 of 6, a legal obligation 2 of 6, not stated 1 of 6 (6 standards, read 8 October 2026).

Asks where the data came from, by standard: asks for the data's sources or provenance 4 of 6, not stated 2 of 6 (6 standards, read 8 October 2026).

Asks about personal data, by standard: not stated 4 of 6, asks whether personal data is included 2 of 6 (6 standards, read 8 October 2026).

What it asks the card to contain, by standard: stated 6 of 6 (6 standards, read 8 October 2026).

Reference: "What each data card standard asks a dataset to document", updated 2026-10-08, https://datacardgenerator.com/datasets/datacardgenerator-rules-from-primary-source/.

Embed code, no JavaScript

6 published standards and legal templates say what a dataset's documentation must contain, each quoted from its own text, read 8 October 2026.

What each data card standard asks a dataset to document, datacardgenerator.com

<figure style="margin:0;padding:16px;border:1px solid #d0d0d0;background:#ffffff"><blockquote style="margin:0 0 8px;background:#ffffff">6 published standards and legal templates say what a dataset's documentation must contain, each quoted from its own text, read 8 October 2026.</blockquote><figcaption style="background:#ffffff"><a href="https://datacardgenerator.com/datasets/datacardgenerator-rules-from-primary-source/" style="background:#ffffff">What each data card standard asks a dataset to document, datacardgenerator.com</a></figcaption></figure>

For machines: JSON, CSV, Markdown.

Save the CSV (44 answers, compiled 2026-10-08)