Provision record
Hugging Face · Hugging Face Model Card Guidelines · View original document ↗

Model card must disclose training datasets

Medium severity Medium confidence Explicit document language Common · 273 of 352 platforms

Key Facts

What must the model card disclose?
Hugging Face requires that the model card disclose which datasets were used to train the model.
Must the model card disclose which datasets were used to train the model?
Hugging Face requires that the model card disclose which datasets were used to train the model.
Stay ahead of the changes
Track Hugging Face and get the diff the day its terms change.
Share 𝕏 Share in Share 🔒 PDF

This analysis describes what Hugging Face's agreement states, permits, or reserves. It does not constitute a legal determination about enforceability. Regulatory applicability and practical outcomes may vary by jurisdiction, enforcement context, and individual circumstances. Read our methodology

ConductAtlas Analysis

Why it matters (compliance & governance perspective)

Mandatory dataset disclosure means model creators cannot omit training data provenance from the model card, which is consequential for transparency and downstream users assessing data lineage.

Interpretive note: The excerpt is a fragment and likely one item in a broader list of model card requirements. The claim is grounded in the verbatim language but the surrounding context is not available.

Clause Stability Stable

0
Changes
3
Months Monitored
Jul 10, 2026
First Seen
Jul 10, 2026
Last Seen
This clause type exists across 1423 other provisions on other platforms.

Consumer impact (what this means for users)

As a model creator, you must identify the training datasets in your model card.

How other platforms handle this

Grubhub Medium

your Review must clearly disclose any "material connection" you may have with Grubhub or the Merchant

Uber Medium

Any person nominated or appointed as arbitrator to a dispute initiated under this Agreement shall disclose any previous or existing relationship with a Litigation Funder...This obligation is continuous and requires ongoing disclosure at any time while the arbitration is pending.

Twitch Medium

You will display or read out the following disclaimer when promoting, administering, or conducting a Promotion: "This is a promotion by [Your Name]. Twitch does not sponsor or endorse this promotion and is not responsible for it."

See all platforms with this clause type →
▸ View Original Clause Language DOCUMENT RECORD
"
which datasets were used to train your model

Excerpt from Hugging Face's Model Card Guidelines

Provision details

Document information
Document
Hugging Face Model Card Guidelines
Entity
Hugging Face
Document last updated
May 12, 2026
Tracking information
First tracked
July 9, 2026
Last verified
July 9, 2026
Record ID
CA-P-064778
Document ID
CA-D-00842
Evidence Provenance
Source URL
Wayback Machine
Content hash (SHA-256)
761f69a9d232f03c9a5687a54a0727406e3e8fd9f091bdb8e1cc1239e4ab3ce9
Analysis generated
July 9, 2026 06:06 UTC
Methodology
Evidence
✓ Snapshot stored   ✓ Hash verified
Citation Record
Entity: Hugging Face
Document: Hugging Face Model Card Guidelines
Record ID: CA-P-064778
Captured: 2026-07-09 06:06:03 UTC
SHA-256: 761f69a9d232f03c…
URL: https://conductatlas.com/platform/hugging-face/hugging-face-model-card-guidelines/provision/CA-P-064778/model-card-must-disclose-training-datasets/
Accessed: Aug. 19, 2026
Permanent archival reference. Stable identifier suitable for legal filings, compliance documentation, and research citation.
Classification
Severity
Medium
Categories

Other risks in this policy

Get the research letter

Companies change their terms quietly. We read every version and catch what actually changed. One email a week on the changes that matter and what they mean.

Frequently Asked Questions

What does Hugging Face's Model card must disclose training datasets clause do?

Mandatory dataset disclosure means model creators cannot omit training data provenance from the model card, which is consequential for transparency and downstream users assessing data lineage.

How does this clause affect you?

As a model creator, you must identify the training datasets in your model card.

How many platforms have this type of clause?

ConductAtlas has identified this type of provision across 273 platforms. See the full comparison.

Is ConductAtlas affiliated with Hugging Face?

No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by Hugging Face.