EU Commission releases GPAI training data template for AI Act compliance

The European Commission has unveiled a standardized template designed to assist General-Purpose AI (GPAI) providers in summarizing the datasets used to train their models. This initiative directly supports the transparency obligations set out in the AI Act, which will become applicable across the European Union from 2 August 2025. The template offers a straightforward and uniform approach for providers to publicly disclose key information about the data underpinning their AI systems.

Executive Vice-President Henna Virkkunen emphasized that the adoption of this template marks a significant step toward fostering trustworthy and transparent artificial intelligence in the EU. The document is intended to facilitate compliance with the AI Act by simplifying the process for GPAI providers to report on their training data, thereby enhancing public trust and supporting the responsible deployment of AI technologies for societal and economic benefit.

Given the vast and often opaque nature of data used in training general-purpose AI models, the template requires providers to detail the main data collections and describe other sources utilized. This increased transparency will not only inform the public but also assist stakeholders with legitimate interests, such as copyright holders, in safeguarding their rights under EU law.

The release of this template forms part of a broader regulatory framework for general-purpose AI models, complementing recent EU guidelines on the scope of the rules (published 18 July) and the General-Purpose AI Code of Practice (released 10 July).