Europe can't just write the rules of AI
The US and China build the models, Europe writes the laws. Rules matter, but alone they aren't enough: where I think Europe (and Italy) can really play its game.
Mattia PanattaUpdated 12 October 20266 minOpinion
A joke has been doing the rounds for years: “America innovates, China copies, Europe regulates”. Like all jokes it’s unfair, but open the Artificial Analysis leaderboard and you’ll see why it stings.
The numbers don’t lie
Among the most capable models in October 2026 you find Anthropic, OpenAI, Google and Meta (United States) and, among open-weights ones, Xiaomi, Z AI, Moonshot, Alibaba and DeepSeek (China). The first European model on the list, Mistral Large 4, sits much lower, and Mistral sells it at a promotional price to win space. Infrastructure shows the same gap: according to the International Energy Agency, in 2024 the US used 45% of the world’s data-centre electricity, China 25%, Europe 15%.
Meanwhile Europe wrote the world’s most comprehensive AI law, and in 2026 had to postpone a major part of it because businesses weren’t ready.
Rules matter (really)
I’m not among those who think the AI Act is a mistake. The risk-based logic is right: a chatbot writing emails shouldn’t be treated like software deciding whom to hire or lend to. The duty to say content is generated, or that you’re talking to a machine, protects everyone. And a clear rule, in the long run, helps businesses too: they know what they can do.
The problem isn’t the rules. It’s thinking rules are enough.
Where Europe can play
Competing for the world’s biggest model takes hundreds of billions in investment, abundant energy and a capital market Europe doesn’t have today. But the AI game isn’t played only there. I see at least four fields where we can win.
1. Adoption. AI’s economic value lies not only with those who build models but with those who use them best. A fabric of SMEs like Italy’s, if it learns to use AI in daily processes, can gain more productivity than it loses by not having a “national champion”.
2. Open, specialised models. You don’t need the world’s best model to read invoices, analyse contracts in Italian or assist a GP. Open models adapted to European languages, rules and sectors, running on European infrastructure, are a concrete path.
3. Trust as a product. A European company selling compliant AI, with data staying in Europe and verifiable guarantees, has an argument American and Chinese competitors struggle to match. Here rules become a competitive advantage.
4. Skills. European researchers are behind many breakthroughs of modern AI, but often work in American labs. Keeping and attracting talent is worth more than any industrial plan.
And Italy?
Italy was the first European country with a national AI law. Good. But Italy’s real challenge is different: millions of small businesses and professional firms without an IT department, at risk of falling behind not for lack of tools (they’re available to everyone for €20 a month) but for lack of time, training and confidence.
That’s why I created Lumen: because in Italy the difference will be made by those who use these tools with method, not by those who build them.
In one line
Rules protect, but they don’t produce. Europe must write the laws of AI and, at the same time, become the place where AI is used best.
Sources
Informational content, current as of the date shown. Prices, models and rules change often: always check official sources. Written with the support of AI tools and reviewed by the author.
