This so-called ‘transparency template’ is nothing more than a smokescreen. It fails to deliver the meaningful disclosure the law demands – and rightsholders deserve.”

News Media Europe (NME), representing leading news publishers across the EU, today expressed its profound disappointment in the European Commission’s finalised transparency template for general-purpose AI models under Article 53 of the AI Act.

While the AI Act rightly obliges providers to publish a “sufficiently detailed” summary of training data, the Template released today is alarmingly superficial. It lacks the specificity and granularity necessary for rightsholders to verify whether their copyright-protected content has been exploited — let alone to enforce their rights effectively.

“This Template lets model developers off the hook. Instead of shining a light on what content was used, it wraps the process in vague ‘narrative summaries’ and cherry-picked domain name lists,” said Wout van Wijk, Executive Director of News Media Europe. “The Commission had the opportunity to set a global precedent for AI transparency. Instead, it blinked.”

Despite extensive consultations, the final outcome reflects the interests of major AI providers far more than those of Europe’s creative sector. The reliance on broad categorizations, undisclosed datasets, and narrative generalities makes it nearly impossible to determine whether protected journalistic content was used – especially content hidden behind paywalls or accessed in violation of scraping rules.

A Betrayal of the AI Act’s Intent

The AI Act – and particularly Recital 107 – clearly states that summaries should empower rightsholders to exercise and enforce their rights. Yet the current Template enables opacity, not transparency. Requiring providers to reveal only the “top 10%” of scraped domains while shielding the majority of datasets from scrutiny renders enforcement a legal guessing game.

“We’re expected to defend our rights against black-box models trained on undisclosed data from undisclosed sources,” van Wijk continued. “This is not compliance. It’s complicity.”

The Template even allows companies to sidestep disclosure entirely by invoking trade secrets — a loophole so wide, it swallows the rule. And while the Commission hints at optional disclosure “upon request,” this voluntary gesture is a poor substitute for the mandatory transparency the AI Act promises.

News Media Europe Calls for Immediate Revisions

News Media Europe calls on the European Commission to urgently revise the Template and establish an enforcement mechanism with real teeth — including:

• Mandatory disclosure of all scraped domains, not just a curated sample.
• Itemised identification of licensed versus unlicensed datasets.
• A binding “upon request” mechanism with deadlines and penalties.
• Clear consequences for providers who fail to comply in good faith.

The Commission must uphold the integrity of the AI Act and stand with European creators, not the tech giants who profit from their work without consent, attribution, or remuneration.

“Europe’s media is not a free training set. Transparency is not a favour — it’s a right enshrined in law,” van Wijk concluded.

Contact: Wout van Wijk