“Don’t be like Europe” has pretty much become the reflexive warning in AI development. It is what a founder tells a room of investors and what a minister tells a room of allies, shorthand for a continent that loves writing the rules of industries it desperately struggles to build.
JD Vance said as much at the Paris AI summit, warning that “excessive regulation of the AI sector could kill a transformative industry.” The EU’s own Draghi report on competitiveness all but conceded the point.
There is a lot of talk about what Europe gets wrong and almost none about the people who wrote the rules in the first place. Lucilla Sioli, Henna Virkkunen, Wojciech Wiewiórowski, Axel Voss. These are the names we should know.
So who are the people building Europe’s AI rulebook? Let’s have a look.
Who Is Lucilla Sioli and What the EU AI Office Actually Does
Every rule needs an enforcer, and the EU AI Act’s is a body that did not exist three years ago.
The European AI Office sits inside the Commission and polices the general-purpose models under ChatGPT, Gemini and Claude. It can demand to see how a model works, order changes, and fine a company up to three percent of its global turnover.
Running it is Lucilla Sioli, an Italian economist who went from industrial policy to holding the leash on the most powerful AI on earth.
Whether that leash is real is the open question. The office opened with around 140 staff, up against companies that spend tens of billions a year and hire away the very experts it needs to referee them.
Its first test was pretty big. Meta refused to sign the office’s voluntary code, and its policy chief Joel Kaplan declared that Europe was “heading down the wrong path on AI”.
Henna Virkkunen and the Digital Omnibus That Softened the AI Act
If you want to know whether Europe will actually punish American tech, look at what Henna Virkkunen did to Elon Musk.
Last December the Finnish vice-president, whose portfolio carries the faintly imperial title of Tech Sovereignty, Security and Democracy, was the face of a €120 million fine against Musk’s X, the first penalty ever issued under the Digital Services Act.
So it is a surprise that the same politician spent this year making Europe’s AI rules easier to live with. Her simplification drive produced the Digital Omnibus, sold as a way to clear “regulatory clutter”, and it was the biggest rewrite the EU AI Act has had since it passed. It pushed the hardest obligations, covering hiring, biometrics, critical infrastructure and policing, from 2026 out to December 2027. And in the same breath it banned nudifier apps, the tools that generate fake intimate images. It stayed hard on the apps nobody would defend and went easy where industry had been pushing.
When Parliament’s civil-liberties committee took evidence in January, MEPs asked outright whether US lobbying had shaped the drafting, and Axel Voss, one of the Parliament’s longest-standing voices on the Act, has kept pushing to enforce it hard rather than let it blur into vagueness. The woman who fined Musk without blinking had, on AI, handed industry much of what it asked for.
Who Enforces the EU AI Act Across the 27 Member States
Sioli and Virkkunen are the headline acts in enforcing the EU’s AI regulation. The rest of the machine is messier, and it is where “Brussels regulates AI” stops being quite true.
The AI Office handles the big models. Everything else, the hiring filters, the credit-scoring engines, the border cameras, falls to national regulators, one or more in each of the 27 member states.
Wojciech Wiewiorowski, the European Data Protection Supervisor, covers the AI the EU institutions use on themselves. A European AI Board is meant to keep all of them in line, so a system blocked in Dublin is not nodded through in Warsaw.
The catch is that a year in, several countries still had not named the regulator meant to enforce the law on their soil. It is a strange kind of overregulation that forgets to hire the regulators, and it is the reason the industry welcomed the delay.
Why the EU Delayed the AI Act’s High-Risk Rules
It is tempting to say that Digital Omnibus is proof Europe is going soft. The more likely version is more awkward, because the case for the delay is strong.
The hard rules were postponed partly because the tools to obey them did not exist. The technical standards firms were meant to build against, the checklists that turn a legal principle into an engineering test, were unfinished and not due until late 2026.
Enforcing a rule before anyone has written down what passing looks like would not have survived a day in court. On that reading the Omnibus is a sequencing fix rather than a surrender, and the obligations were delayed rather than dropped.
Delay does have a habit of settling in, and every month the rules sit unused the systems they govern dig in deeper. Both can be true. Sioli’s office can be right on the law and still be losing the argument about who sets the pace.
What EU AI Act Enforcement Looks Like Now
The real test of who regulates AI in Europe is not in the text of the Act. It is in what Sioli’s office does next.
Meta is the first fight that counts. A company is betting in public that Brussels lacks the money and the nerve to punish a refusal, against a regulator that needs an early win to prove the law bites. Whoever is right sets the tone for everyone after.
Then watch where the fight moves. The Commission has already opened a second front with the Cloud and AI Development Act. The EU will continue to fight the “Don’t be like Europe” narrative, but whether it succeeds or not is what we will be watching next.
Author: Grace Sharp
See Also:
What Did the EU’s Digital Omnibus Actually Delay in the AI Act?
What Is an Autonomous AI Attack, and Has One Actually Happened?
