Following the publication of the so-called “transparency template” by the European Commission’s AI Office on 24 July, most AI model providers will be required to publish summaries of training data as of 2 August 2025. Although these templates were initially designed to facilitate copyright enforcement by obliging AI developers to disclose the origin of their training data, the European Federation of Journalists (EFJ) joins the media and creative sectors in criticising the finalised template, having repeatedly called for more detailed transparency obligations.
Required by the AI Act, this obligation for AI model providers to disclose the data used to train algorithms could have been a useful instrument for rightsholders to prove AI developers have stolen their works. However, the template requires providers to reveal only the top 10% of scraped domains, while allowing companies to circumvent their obligations by invoking trade secrets.
The EFJ, together with the German journalists’ association DJV, strongly advocated for maximum transparency to protect journalistic pieces and help authors take legal action against the unauthorised and unpaid use of their work.
“The EU Commission has finally caved in to the big tech companies,” says DJV Managing Director and AI specialist Hanna Möllers. “The AI Office is giving authors stones instead of bread, and is disregarding the will of the European Parliament as set out in the AI Act.”
Möllers also criticises the Code of Practice for General Purpose AIs, published two weeks ago, a code of conduct that was actually intended to further strengthen compliance around key issues such as copyright. But here, too, the final text falls short of what rights holders expected, namely a tool to counter unfair AI practices and hope to effectively enforce its rights.
On 30 July, a coalition of European creators, including the EFJ, further criticised the outcome, denouncing months of negotiations as a failure to address the sector’s core concerns, instead favoring the interests of major AI providers: “We strongly reject any claim that the Code of Practice strikes a fair and workable balance or that the template deliver sufficient transparency about the majority of copyright works or other subject matter used to train GenAI models. This is simply untrue and is a betrayal of the EU AI Act’s objectives,” the statement reads.
The EFJ joins the creative sector in calling on the European Commission to review the implementation package and meaningfully enforce Article 53, and on the European Parliament and Member States to challenge the unsatisfactory process around the Code of Practice and transparency templates.
Source: EFJ
