Unmasking AI: Making the EU AI Act Succeed – CEPIC’s Key Takeaways from Brussels 2025

CEPIC wants AI to succeed for everyone. The whole value chain should benefit. In addition, rights holders should be able to opt out. Effective opt-out mechanisms should be in place. The template developed by the AI Office plays a crucial role in achieving this. While negotiations on the implementation of the AI Act are taking place in Brussels, seven organisations representing press, book and scientific publishers, music publishers and visual content (CEPIC) have organised an industry event at the Press Club, next to the European Commission and the Council.

Organisations sponsoring the Event: ICMP ; CEPIC ; EPC ; FEP ; ENPA ; EMMA ; RELX

On March 10, 2025, industry leaders, policymakers, and copyright experts gathered in Brussels for Unmasking AI – Making the EU AI Act Succeed. The event focused on the growing impact of AI-generated content, the challenges surrounding copyright protection, and the urgent need for legal frameworks that ensure fairness, transparency, and accountability in AI development.

CEPIC was proud to be represented by Patrick Lacroix, CEO of Belga, who spoke about the critical issues facing the visual media industry in the wake of rapid advancements in generative AI. His speech underscored one undeniable fact: without full transparency, AI will continue to undermine creators, disrupt licensing models, and erode trust in visual content.

Lacroix’s speech began with a stark example—the viral, AI-generated image of Pope Francis in a white puffer jacket. Millions believed it was real, highlighting how AI-generated content, when unchecked, has the power to mislead and manipulate public perception. The core question remains:

• Who controls AI-generated content?
• Who is responsible when AI misleads?
• How do we create a system where AI and copyright can coexist fairly?

‍

The Visual Industry’s Stance: AI can be a tool but only with legal certainty

The visual media industry is not opposed to AI, in fact, it has the potential to be a powerful tool for creators. However, the lack of transparency and proper legal frameworks makes it impossible for content creators to trust or fairly benefit from AI-generated works.

Currently, Belga and many other news agencies choose NOT to use AI-generated images. Not because they reject AI, but because:

• There is no way to verify whether an AI-generated image was created using copyrighted content.
• The origin of AI training data is hidden, making it impossible to track whether a dataset has lawfully obtained images.
• No clear framework exists for how AI and licensing should work together.

Without transparency, AI cannot be trusted as a tool for the visual industry.

‍

The AI Act’s Transparency Requirements: Why they must be strengthened

At the heart of the discussion was the AI Act’s mandatory transparency template (Article 53), which is intended to provide clarity on what datasets AI companies use. However, the current version of this template is insufficient.

Major flaws in the current transparency template:

• It only requires AI companies to disclose "major sources" which is meaningless for copyright holders.
• Scraped URLs are not disclosed, making it impossible for creators to see if their work was used.
• The 5% threshold excludes small players, meaning independent photographers, illustrators, and agencies may never know if their content was used.

AI companies claim that revealing training datasets would expose "trade secrets"—but this is false. The pharmaceutical industry discloses the ingredients of medicines without giving away formulas. The same logic must apply to AI: disclose the datasets, not the technology behind them.

‍

Why the AI Act Must Deliver Real Protection

One of the biggest concerns discussed was how AI companies have used the Text & Data Mining (TDM) exception to justify mass scraping of copyrighted works.

• AI scrapers ignore opt-out mechanisms like robots.txt and metadata blocks.
• Instead of strengthening protections, rightsholders now invest in content poisoning tactics, deliberately corrupting datasets to make AI training more costly.
• This is NOT a sustainable AI ecosystem, it’s a war between creators and AI developers.

The solution?

• ‍Move from “opt-out” to “prior authorisation.” AI companies should not have the right to scrape content without explicit permission.
• Implement strong compensation mechanisms. Rightsholders must have a way to monetise their content, even if it was already used in AI models.
• ‍Create an effective enforcement system. There should be a central regulatory body to ensure AI companies comply with copyright laws.

AI and Copyright Can Coexist—But Only with Fairness and Transparency

CEPIC and the photography industry are not against AI but we do demand legal protection, fair compensation, and full transparency to ensure AI supports rather than undermines the visual media sector.

What must be included in the final AI Act:

‍
🔹 Full disclosure of all AI training datasets—not just vague references to "major sources."
🔹 Transparency on synthetic AI training—AI models trained on AI-generated content should still be traceable.
🔹 No exemptions for SMEs—AI impact isn’t about company size. Transparency must apply to all AI companies.
🔹 A simple, digital opt-out system for creators, along with strong enforcement mechanisms.
🔹 AI companies must disclose how they use copyrighted works, including resale, licensing, or generating derivative content.

‍

The visual industry is ready to embrace AI but only under a legal framework that protects rightsholders, ensures fair compensation, and prevents AI from becoming a tool for unchecked copyright infringement.

As Patrick Lacroix stated in his closing remarks:

‍"AI can be part of the visual industry’s future—but only if we get it right. Without transparency, rightsholders lose control, AI models degrade, and legal uncertainty stifles innovation. The AI Act is Europe’s moment to fix this. We must demand full transparency, clear compensation, and the right to control how our works are used."

‍

CEPIC will continue advocating for these essential changes to the AI Act. Stay tuned for updates as we push for a fair and sustainable future for AI and the visual media industry