AI Governance Institute logo
AI Governance Institute

Practical Governance for Enterprise AI

VoluntaryGuidelineEU

EU General-Purpose AI Model Training Data Public Summary Template

Issued by

European Commission

liveEffective 2026-01-01GPAI-TDSVerified April 2026

The European Commission published a standardized template for providers of general-purpose AI models to use when publicly disclosing summaries of their training data. It supports compliance with the transparency obligations for general-purpose AI models established under the EU AI Act. Providers are expected to follow the template structure when meeting their disclosure requirements under that regulation.

Applies To

Large enterpriseSMBAI developer

Overview

This template is an implementation instrument issued by the European Commission to operationalize the training data transparency requirements applicable to general-purpose AI model providers under the EU AI Act. The EU AI Act requires providers of general-purpose AI models to make publicly available a sufficiently detailed summary of the content used to train the model, enabling third parties to assess potential copyright, privacy, and bias considerations. The template standardizes the format and minimum content of those summaries, reducing ambiguity for providers and creating a consistent basis for regulatory review. It applies to all providers placing general-purpose AI models on the EU market, regardless of where those providers are established. While styled as a guideline, the template functions as a quasi-binding compliance tool because adherence is expected as evidence of satisfying the underlying statutory obligation. The European Commission may revise the template as enforcement practice and technical understanding of training data documentation matures.

Key Requirements

  • Providers of general-purpose AI models must publish a training data summary using the Commission-issued template structure
  • Summaries must disclose, at minimum, the main data sources, data types, and any filtering or selection criteria applied during training
  • Disclosures must be made publicly available, meaning accessible without restriction to any third party
  • The template must be completed prior to or at the point of placing a general-purpose AI model on the EU market
  • Failure to comply with the underlying EU AI Act transparency obligation can result in fines of up to EUR 15 million or 3% of global annual turnover, whichever is higher
  • Providers with systemic-risk-designated models face additional and more detailed disclosure obligations beyond the baseline template

What Your Organization Must Do

  • Appoint a designated compliance lead by Q3 2025 to own the training data disclosure process and coordinate across legal, data engineering, and product teams ahead of the January 1, 2026 effective date.
  • Map all general-purpose AI models intended for the EU market and confirm whether any have been designated as systemic-risk models, as these require more detailed disclosures beyond the baseline template.
  • Complete the European Commission training data summary template for each in-scope model before or at the point of EU market placement, ensuring disclosures cover main data sources, data types, and filtering or selection criteria at minimum.
  • Publish completed summaries in a publicly accessible location with no access restrictions, and document the publication date and URL in your compliance records to support regulatory review.
  • Conduct an internal review of existing training data documentation against the template structure now, identifying gaps in data provenance records that data engineering teams must remediate before the deadline.
  • Establish a monitoring process to track European Commission revisions to the template and trigger an update cycle for published summaries within 90 days of any new version release, given that fines for non-compliance with the underlying EU AI Act obligation can reach EUR 15 million or 3% of global annual turnover.

Playbook Guidance

Step-by-step implementation guidance for compliance teams.

Frequently Asked Questions

Is the GPAI training data summary template legally binding or just a recommendation?
The template is styled as a guideline but functions as a quasi-binding compliance tool. Adherence is expected as evidence of satisfying the transparency obligations under the EU AI Act. Non-compliance with those underlying obligations can trigger fines of up to EUR 15 million or 3% of global annual turnover, whichever is higher.
What information must be included in a GPAI training data summary under this template?
At minimum, providers must disclose the main data sources, data types, and any filtering or selection criteria applied during training. The summary must be publicly accessible without restriction, enabling third parties to assess copyright, privacy, and bias considerations.
Does the GPAI training data summary requirement apply to AI model providers based outside the EU?
Yes. The requirement applies to all providers placing general-purpose AI models on the EU market, regardless of where the provider is established. Non-EU companies must complete and publish the template before or at the point of EU market placement.
When does the GPAI training data summary template take effect and what is the compliance deadline?
The template becomes effective January 1, 2026. Providers should complete the template for each in-scope model before placing it on the EU market. Organizations are advised to appoint a compliance lead and audit existing training data documentation well before that date.
Do providers of systemic-risk GPAI models have additional disclosure obligations beyond this template?
Yes. Models designated as systemic-risk models under the EU AI Act face more detailed disclosure requirements beyond the baseline template. Providers should confirm whether any of their models carry a systemic-risk designation and prepare for the additional documentation obligations that apply.
What happens if the European Commission revises the GPAI training data summary template after initial publication?
The Commission may revise the template as enforcement practice and technical understanding mature. Providers should monitor for updates and plan an internal update cycle for already-published summaries, since continued compliance with the underlying EU AI Act obligation depends on using the current template version.
EU General-Purpose AI Model Training Data Public Summary Template — Overview & Implementation Guidance | AI Governance Institute