Flanders Minister of Mobility, Annick De Ridder (NV-A), has published the first safety data from three months of Tesla’s supervised self-driving system on Belgian roads. The encouraging results report only five minor incidents. At the same time, FSD had its deadliest month in the US. 

Since June, owners of Teslas equipped with FSD are allowed to deploy the self-driving feature in Flanders. The approval has already harvested some interesting data, as these cars covered roughly 24.6 million kilometers, according to figures presented by mobility minister Annick De Ridder

Small scrape

Between 11 June and 31 August, only five incidents were registered, and only one was formally reportable: a minor scrape in which the driver had reportedly disengaged the automatic pilot shortly before impact. 

The other four were what the minister described as “very light accidents”, like a bicycle handle brushing a door. None of these were legally reportable, and most involved third parties.

For comparison, non-self-driving Teslas covered 147 million kilometers in the same period and were involved in 150 reportable collisions. De Ridder concluded that off motorways, self-driving Teslas carried almost ten times lower accident risk, while on motorways the reduction was roughly 30%.

Official vote

These early data suggest the first trials are going well. Machines are outperforming humans, which underpins the technology. These findings are crucial, as Europe is preparing an official vote on the 6th of October on the broader approval of FSD, which is now admitted in several member states under a loophole.

The Scandinavian countries urge a vote against, while bigger nations like France and Germany, which play a decisive role in the vote, remain cautious.

But against the promising early data from Belgium and the Netherlands, another, grimmer picture of FSD is emerging. In July, the American NHTSA posted the highest number of FSD accidents ever recorded in a month. 

In total, 236 FSD-related crashes were reported. Four of them were fatal, with seven people losing their lives.

Two sides to the problem

In Maryland, a Tesla Model 3 making an unprotected left turn failed to yield to a motorcycle, killing three people. In San Mateo, a Model S left the road, hit a light pole and caught fire, killing the driver. In Ione, a Model Y crossed the center line and hit another vehicle head-on, killing its 25-year-old driver. Only a crash in Minnesota was different, where a collision in a roadworks zone involved a Ford reportedly traveling the wrong way. But FSD was active on the Tesla.

The problem is two-fold. The statistics show that the number of FSD-related accidents in the US is rising, from 82 in the same month last year to almost triple that number this year. If the automated system improves by learning from more mileage, the numbers paint a different picture.

Also, Tesla doesn’t release data on these incidents to protect its business case. That same argument is used to explain why it didn’t publicly share test data before approval in the Netherlands, which snowballed the rollout. As a result, the system’s precise involvement is hard to pinpoint. 

In any case, almost no other car maker is so secretive about its autonomous-driving data.

Different systems

Then again, the American and European versions of FSD are not directly comparable. Not by software, and not by rule.  

In the U.S., Tesla operates under a self-certification regime with post-market oversight. The software includes more assertive driving profiles. The European configuration uses stricter continuous driver monitoring via cabin cameras, requires hands to remain available for immediate takeover, and defaults to more conservative driving behavior. For instance, if the driver looks at a phone, the system issues warnings and ultimately shuts down.

The Belgian trial works because European regulators have refused to let Tesla use public roads as a laboratory. The latest American statistics show why that distinction matters.