Europe stands at a crossroads on artificial intelligence and copyright, with new regulations threatening to either protect creative workers or stifle homegrown AI innovation, depending on how lawmakers decide. As the European Union prepares to revise its copyright rules, a heated debate is unfolding over how AI systems should be allowed to train on text and data, and whether creators deserve compensation when their work fuels these systems.

What's Driving the Copyright-AI Conflict in Europe?

The tension centers on a fundamental question: when an AI company trains a large language model (LLM), a type of artificial intelligence that learns patterns from vast amounts of text to generate human-like responses, should it pay authors, artists, and publishers whose work was used? The European Union's Copyright Directive, originally updated in 2019, didn't anticipate the explosion of generative AI. Now, as the directive faces review, the European Commission is asking stakeholders for feedback on creative content licensing for AI training to ensure proper remuneration for creators in today's digital age.

On May 27, 2026, the AI Chamber convened a roundtable discussion in Brussels bringing together policy experts, industry leaders, and European Commission representatives to tackle this exact problem. The conversation revealed deep uncertainty about how existing copyright rules interact with the EU's new AI Act, a comprehensive regulatory framework that took effect in 2024.

The core issue is legal ambiguity. Articles 3 and 4 of the Copyright Directive address text and data mining, a process where AI systems automatically extract patterns from large volumes of written material. But the rules were written before modern AI systems became commonplace, leaving companies unsure whether their training practices are legal.

How Are European AI Companies Being Affected by This Uncertainty?

The practical consequences are already visible. Some European startups and research institutions are relocating operations abroad to avoid legal risk, according to roundtable participants. This brain drain threatens Europe's ability to compete globally in AI development.

The stakes are particularly high for smaller players. European startups, small and medium-sized enterprises (SMEs), and research institutions need to train their own foundational models, the large AI systems that power downstream applications. But the legal uncertainty surrounding copyright makes this expensive and risky. If companies must negotiate licenses with every copyright holder whose work they use, the costs could become prohibitive, putting European developers at a disadvantage compared to competitors in jurisdictions with clearer rules.

One concrete example emerged during the discussion: practical examples of how this legal complexity has already impacted European projects attempting to train large language models were presented, though the specific projects were not named.

Steps for Policymakers to Navigate the Copyright-AI Balance

• Clarify Legal Uncertainty: The European Commission must provide clear guidance on whether text and data mining for AI training is permitted under existing copyright law, or whether new rules are needed to define the boundaries.
• Avoid Prohibitive Licensing Requirements: Any new rules should not impose mandatory licensing schemes that would make it economically impossible for European startups to train competitive AI models without paying unsustainable fees.
• Protect Creator Compensation: Establish a balanced framework that ensures authors, artists, and publishers receive fair compensation when their work is used to train AI systems, without creating barriers to innovation.

Interestingly, even copyright holders themselves are divided on the solution. During the roundtable, participants noted that there is no clear consensus on the rightsholder side about whether legislative intervention is necessary at all, or whether letting the market establish an equilibrium is the wiser choice.

The European Commission's approach matters enormously. The AI Act itself may give European copyright law extraterritorial reach, meaning that rules set in Brussels could affect how companies worldwide train AI systems. This amplifies the importance of getting the balance right.

The timing is critical. The European Commission is expected to publish its tech sovereignty package on June 3, 2026, which will include proposals aimed at strengthening Europe's digital infrastructure. Separately, the review of the 2019 Copyright Directive is expected to produce a legislative proposal early next year, with the European Commission asking stakeholders to submit feedback now.

"We need to balance intellectual property protection with the practical, economic realities of text and data mining so that European startups, SMEs, and research institutions can thrive without prohibitive friction and costs that would put them at a disadvantage with peer competitors from different jurisdictions," the AI Chamber stated after the roundtable discussion.

AI Chamber CEE, Policy Organization

The European Commission's Rodolphe Wouters participated in the May 27 discussion, signaling that Brussels is actively listening to stakeholder concerns. The EU business community is now eagerly awaiting the results of the stakeholder survey and the copyright directive review process.

What happens next will shape whether Europe can build world-class AI companies or whether the continent becomes dependent on non-European foundational models. The copyright question is not abstract; it is a practical lever that will determine whether European innovation can compete globally or whether regulatory friction pushes talent and investment elsewhere.