Germany’s 2021 Autonomous Driving Act was, at the time, the most permissive L4 regulatory framework in the world. The country pushed for a Europe-wide equivalent, and got one. Type approval for fully autonomous vehicles is now a defined pathway in the EU regulatory acquis. The technical specifications are public. The conformity assessment bodies are accredited. The paperwork exists.

What does not exist, two years in, is a queue.

Mercedes-Benz Drive Pilot is the only L3 system with EU type approval on a passenger car. There are no certified L4 vehicles on sale. The handful of L4 testing programs that did file, Mobileye in Munich, Stellantis with an industrial partner in northern Italy, the smaller Continental and Bosch research fleets, have all stalled at the conformity-assessment stage, where the homologation bodies are asking the same question: what is the operational design domain you want us to approve, and how have you proven the vehicle stays inside it?

This is the question Europe’s regulatory pathway was supposed to be good at answering. It is, on paper, the question the framework was designed around. In practice, the framework asks for evidence the applicants either cannot produce or cannot produce cheaply enough to be worth it.

Why the door is held open and empty

Three reasons.

The first is that the European market for robotaxi service is, today, small. The candidate cities, Paris, Berlin, Amsterdam, Stockholm, Helsinki, Lisbon, have a combined population that is meaningful but a combined regulatory environment that is fragmented at the city and national level. The type approval gets you the vehicle. The operating permit gets you the city. There is no European-wide operating permit and there is not going to be one in 2026.

The second is that the operational design domain documentation EU regulators are asking for is, in the considered opinion of the engineers who have tried to assemble it, more rigorous than what NHTSA requires. This is not necessarily wrong. It is, however, expensive, and it is more expensive than the equivalent US filing for what is, from an investor perspective, a smaller addressable market.

The third is that the Tier-1 suppliers who would normally do this work, Bosch, Continental, ZF, Valeo, Aptiv, are not, in 2026, in the business of deploying their own L4 fleets. They sell into OEM programs. The OEMs are not, in 2026, deploying L4 fleets. The buyer is missing.

What changes the math

Two things, both of them somewhat unrelated to the regulatory framework itself.

If a Chinese manufacturer with a working L4 stack files for EU type approval on a vehicle they want to sell to European fleet operators, the calculus on the conformity assessment changes. Chinese manufacturers have, on the available evidence, the cheapest fully integrated stack in the world. They have, on the available evidence, also been waiting for the European market to define what it wants. If one of them decides the type approval is worth doing, the EU’s L4 regulatory regime gets its first non-research-program filing within twelve months.

The other thing is Mercedes. Drive Pilot is already L3. Mercedes has publicly stated an intention to file for L4 type approval, on a defined operational design domain, before the end of the decade. If the filing goes in, the regulatory machinery has its first reference case and the conformity bodies have their first precedent. Everything that comes after is easier.

Until then, Europe has the most carefully designed L4 regulatory pathway in the world, and almost no one is using it. That is not, on inspection, a failure of the regulation.