European Policymakers and AI: Je T’aime… EUx Non Plus
Brussels has a peculiar brand of rhythm, one that sounds suspiciously like Serge Gainsbourg’s Je t’aime… moi non plus. On one hand, the European Union beats the drum of “digital sovereignty,” professing its love for a homegrown AI ecosystem meant to rival Silicon Valley and Beijing. On the other hand, the European Parliament appears determined to build a regulatory cage so small the bird can’t even hatch, let alone fly.
On January 28, 2026, the JURI Committee took a decisive step into this cognitive dissonance by adopting MEP Axel Voss’s own-initiative report on Copyright and Generative AI. By a vote of 17 to 3, the committee signaled that it prefers the comfort of legacy property models over the messy, necessary reality of technological progress.
The stakes are no longer just academic; they are existential for the EU economy. We are watching a repeat of the early 2000s, where legal uncertainty around web crawling effectively handed the search engine market to foreign giants. Now, we are telling our innovators: “Go ahead and build the future, but do it with your hands tied behind your back and a legal target on your chest.”
From “AI Gigafactories” to Cathedrals in the Desert
The irony of this vote is sharpened by the EU’s simultaneous push for the InvestAI initiative. This program aims to spend €20 billion on building five “AI Gigafactories”—massive data centers housing 100,000+ GPUs each. Commission President Ursula von der Leyen has called them a “CERN for AI.”
However, without access to training data, these facilities risk becoming “cathedrals in the desert”: magnificent physical structures housing rows of empty, silent servers. We are basically spending billions to build the most sophisticated industrial mills in history, while passing laws that make it a crime to mill any grain that wasn’t hand-delivered with a signed certificate of origin. We are subsidising the silos for an industry we are simultaneously outlawing in spirit
The Wikipedia Model: Infrastructure, Not Permission
The JURI report rests on the flawed assumption that AI developers must pay for the right to train under copyright law. But a far more sensible model is already emerging. On January 15, 2026, the Wikimedia Foundation announced API access deals with Microsoft, Meta, Amazon, Perplexity, and Mistral AI.
Crucially, these are not content licensing deals. Wikipedia’s 65 million articles remain freely available under Creative Commons licenses. Instead, these companies are paying for Wikimedia Enterprise: high-speed, high-volume API access tailored for the industrial scale of AI training.
- The Fair Share: As Wikipedia founder Jimmy Wales noted, it isn’t about blocking access to knowledge, but about asking major firms to “chip in” for the massive infrastructure costs their scrapers impose on the foundation.
- The “Access” Market: This proves there is a market for data that doesn’t rely on “shaking down” developers for copyright windfalls. AI companies are willing to pay for better, faster, and more reliable access, not for a “permission to read” that copyright doesn’t actually require.
The Transparency Trap: Counting Grains of Sand
The report’s headline demand for “full transparency” sounds noble in a press release, but it is technically impossible in practice. By calling for an “itemised list” of every single copyright-protected work used for training, the Parliament is demanding that developers count the individual grains of sand in a desert.
Even more concerning is the rebuttable presumption of use. In plain English: If a developer cannot prove they didn’t use a specific workthey are presumed liable, despite the technical impossibility of the “itemised list”.
This isn’t regulation; it’s a litigation trap. Audrey Herblin-Stoop, Mistral AI’s VP of Global Public Affairs, recently warned the French National Assembly that such proposals turn European AI companies into “litigation targets.” This regulatory gap with American and Chinese competitors leaves the EU’s only true global players in an impossible position.
Manufacturing “Ambiguity” in Settled Law
Perhaps the most frustrating aspect of the Voss report is its attempt to manufacture doubt where none exists. The Text and Data Mining (TDM) exceptions (Articles 3 and 4 of the CDSM Directive) were supposed to be the bedrock of European innovation, allowing researchers and companies to mine data unless a rightsholder opted out.
By suggesting this law is “unclear,” the JURI Committee is:
- Reopening a settled debate that the Commission has already confirmed.
- Pushing AI activity into a “grey zone” for the next 2–3 years—the exact window in which global leadership will be decided.
- Inviting market instability just as European startups need certainty to scale.
The “Press Publisher” Privilege: Fairness vs. Property
The report’s push to grant press rightsholders “full control” (requiring explicit consent for every use) is a move to protect legacy business models at the expense of the ecosystem. As legal expert Viktoria Kraetzig argues, this conflict is not a matter of property, but of fairness.
Looking at her paper and the latest market developments, the following remarks can be made:
- Facts Cannot Be Owned: The reporting of facts is publici juris. AI models aren’t “stealing” property; they are processing facts that are already in the public sphere.
- Service vs. License: Wikipedia’s approach shows the way forward. Publishers should charge for enterprise-grade access and utility, not attempt to gate-keep the computational analysis of facts—a move that threatens the very foundations of the open web.
- Creative Destruction: Journalism no longer depends strictly on the traditional platform models of legacy publishers. As Kraetzig notes, EU regulation should aim for a “well-functioning information market,” not a guaranteed income stream for uncompetitive models.
Conclusion: Unrequited Love
Europe is at a crossroads. It can choose to be a “Regulatory Superpower,” exporting intricate rules that offer the comfort of control while suffocating its own innovators. Or it can choose to be a genuine technological power.
By adopting this report, the JURI Committee has doubled down on protectionism over progress. If the plenary follows suit, we should not be surprised when the next generation of AI champions emerges in Austin, Tel Aviv, or Shenzhen.This is the real Je t’aime… EUx non plus of European policy: a declared love for innovation, paired with a regulatory posture that keeps it perpetually at arm’s length. Europe says it wants AI to flourish, but not like this, not here, and not now. In a field where timing and scale decide winners, this hesitation is more than just a pause in the rhythm: it is the sound of an industry being silenced before it has even begun to sing.
Get the article’s core concepts in a unique auditory summary—listen to the AI-generated track below!
This song was created entirely using artificial intelligence tools. If you enjoy this experiment, keep an eye out for each song illustrating our December articles.
Written byCaroline De Cock, LL.M., Head of Research
