FSD Read the 30 Sign and Displayed 50. In Brussels It Averaged 44.

On 24 September 2026 a Brussels road-safety charity published the first independent road test of Tesla's FSD (Supervised) in a European city. It landed twelve days before EU member states vote on whether to let the system loose across the bloc.

Johanna.be, a Brussels non-profit that campaigns for pedestrian and cyclist safety, spent three days in July 2026 covering roughly 400 km, deliberately staying on the low-speed streets where people on foot and on bikes are most exposed. It calls the study FSD30.

What it measured

Fourteen km/h over the limit is not a rounding error on a street designed around the assumption that a child stepping out survives the contact. The 30 km/h zone exists because collision survivability falls away steeply above it.

The display is the more serious finding

The harder problem is not that the car sped. It is that the car appeared not to know it had. In the majority of the 30 km/h zones Johanna.be tested, the screen showed a 50 km/h limit — even though the vehicle had visibly recognised the roadside 30 signs.

That matters because of what "Supervised" means. The driver is asked to catch the system's mistakes, and the speed-limit readout is the instrument a driver would check before deciding whether to intervene. A driver doing exactly what Tesla asks was being shown the wrong number.

The group also recorded FSD repeatedly attempting to overtake cyclists in fietsstraten — bicycle streets where motor traffic must stay behind bikes rather than pass them.

Johanna.be gives the system credit where it earned it: FSD yielded to pedestrians and cyclists, often earlier than it had to.

Why it lands twelve days before the EU vote

The European Commission's Technical Committee on Motor Vehicles votes on 6 October on extending FSD approval to all 27 member states. Passing needs at least 15 countries representing 65% of the EU population. Speeding is not a new objection at that table — it is the specific concern Sweden and France have already raised.

Belgium has now produced both kinds of evidence. On 18 September the Flemish mobility minister presented parliament with its own FSD collision figures: 24.6 million kilometres, five collisions, and a result that flattered Tesla away from motorways. This test measures something those counts cannot. Not whether FSD crashes, but whether it obeys the law on a residential street — and a system that speeds without crashing produces no collision statistic at all.

What it does not show

Four hundred kilometres over three days in one city is a small sample, and Johanna.be is an advocacy organisation with a stated position rather than a road authority. The group did not publish which FSD build it tested, so the results cannot be pinned to a software version or checked against a later one.

None of that disposes of the speed-limit misreading, which is a reproducible software behaviour rather than a statistical claim. Tesla has not publicly responded to the study.