MedTech Terms
    The authoritative reference
    All terms
    Software & AIAI / ML in DevicesSoftware Lifecycle

    Datasheets for Datasets

    Structured documentation describing a dataset's motivation, composition, collection process, preprocessing, uses, and maintenance, the dataset equivalent of a model card.

    Reviewed by Christian Espinosa, Founder, Blue Goat CyberLast reviewed June 20, 2026

    Definition

    Datasheets for Datasets is a documentation framework, introduced by Gebru et al. (2018), that prescribes a structured set of questions every dataset should answer: motivation for creation, composition (instances, labels, splits), collection process (who, how, when, consent), preprocessing/cleaning/labeling, uses (recommended, discouraged), distribution, and maintenance. The intent is to surface biases, gaps, and limitations of a dataset before it is used to train or evaluate a model, and to make those properties auditable downstream by model consumers and regulators.
    What the regulation says
    Datasheets for Datasets is a voluntary documentation approach. IMDRF N88 FINAL:2025 supports representative datasets, independent training and test data, and lifecycle controls, but does not make a specific datasheet template mandatory. Document relevant training, validation, and test data in a form appropriate to the device and submission. This review did not establish that datasheets are the most widely adopted format, widely used in PCCP submissions, or specifically required as separate documents by FDA.

    What this means in practice

    Datasheets for Datasets is a voluntary documentation approach. IMDRF N88 FINAL:2025 supports representative datasets, independent training and test data, and lifecycle controls, but does not make a specific datasheet template mandatory. Document relevant training, validation, and test data in a form appropriate to the device and submission. This review did not establish that datasheets are the most widely adopted format, widely used in PCCP submissions, or specifically required as separate documents by FDA.

    Examples

    • Illustrative: A MedTech company developing an AI for diabetic retinopathy detection uses a datasheet to document the ethnicity and age distribution of patients in its training dataset, highlighting potential underrepresentation of certain demographic groups.
    • Illustrative: Before integrating a new public dataset of medical images into their AI model development, a MedTech firm consults its datasheet to understand how the images were acquired, the specific preprocessing steps applied, and any known biases in the labeling process.
    • Illustrative: A regulatory submission for a high-risk AI medical device includes a datasheet for its validation dataset, detailing the clinical sites where data was collected, patient inclusion/exclusion criteria, and the consent procedures followed.
    Common pitfalls
    • •Documenting only the training set, validation and test set datasheets are equally important for evaluating generalization claims.
    • •Glossing over consent and source legitimacy, these are increasingly material under GDPR, HIPAA, and AI Act scrutiny.
    • •Treating datasheets as static, the document must be updated when datasets are augmented or relabeled.

    Frequently asked questions

    The framework, introduced by Gebru et al. in 2018, prescribes structured questions across motivation for creating the dataset, its composition (instances, labels, and data splits), the collection process including who collected it and how consent was obtained, preprocessing and labeling steps applied, recommended and discouraged uses, distribution terms, and ongoing maintenance responsibility. Answering these questions systematically is what allows downstream model developers and regulators to assess whether a dataset is fit for a given use before it drives training or evaluation.

    Related terms

    Grouped by theme
    Cited by

    Where this term appears across MedTech Terms.

    Sources

    4 sources

    Sources support the definition or relevant regulatory context, not necessarily every claim on this page. Each is graded against our source-tier hierarchy so you can see what rests on binding law versus commentary.

    Official-source claim check:

    Dataset provenance, representativeness, independence, and limitations are relevant to GMLP; a named datasheet format is not established as mandatory.

    Partial review: verified corrections are applied, but this is not a full accuracy certification. The last full review date is unchanged.

    Remaining verification limits
    • Original merged record reviewed in full. Illustrative examples are hypothetical, not verified case histories. Evidence supports the specific propositions below, not an assertion that every linked page verifies all merged content.
    • Original academic framework and consolidated EU AI Act Article 10 applicability and transition dates were not fully verified from primary text in this review. Do not treat the original blanket EU requirement assertion as verified.
    Tier 2Regulator guidance and consensus· 3Tier 3Peer-reviewed and registry evidence· 1
    Link health: 3 verified 1 unchecked· last checked 2026-06-20
    arXiv·1FDA·1IMDRF·2
    1. 1
      Datasheets for Datasets (Gebru et al.)
      Tier 3 Verified
      arXivarxiv.org
    2. 2
      GMLP Guiding Principles
      Tier 2 Verified
      FDAfda.gov
    3. 3
      IMDRF - Software as a Medical Device
      Tier 2 Verified
      IMDRFimdrf.org
    4. 4
      IMDRF N88 FINAL:2025 GMLP
      Tier 2 Unchecked
      IMDRFimdrf.org

    Inline markers like [1] jump to the matching reference above.