As Tesla is gradually but quickly rolling out its supervised autonomous driving tech in Europe, details are emerging about how the company is trying to influence the approval process. According to Reuters, a policy manager is on a roadshow with a polished presentation that overestimates the safety benefits of FSD Supervised.
After Tesla obtained approval in the Netherlands – with Belgium following suit only last week – the company went on a presentation round in Europe to promote FSD Supervised in other member states. The aim was to increase approvals all over Europe.
Independent experts
But the safety data used during that presentation are now being questioned by Reuters, which asked eleven independent traffic-safety researchers to examine the methodology behind those figures. Ten of them concluded that what Tesla had presented to European regulators was not a safety analysis. In their assessment, it was misleading marketing.
The slides, which project safety data amassed in the U.S., were polished. According to Tesla’s representative, Full Self-Driving Supervised could travel more than seven times as far between crashes as the average American driver. It had, the company claimed, the potential to save 32,000 lives annually and prevent 1.9 million injuries.
However, according to the researchers, those claims are structurally invalid. When measuring crashes in FSD-equipped vehicles, Tesla counts incidents that trigger airbag deployments as high-severity events.
It then compares that rate to US federal crash data that covers all collisions requiring a tow truck, which includes far more minor incidents. The two datasets do not measure the same thing.
“What’s your point?”
There is a second distortion. The average Tesla on American roads is 4.1 years old. The average US vehicle is 12.8 years old. Modern cars benefit from mandatory advances in crash-avoidance technology that have nothing to do with autonomous driving. Any comparison between a new vehicle fleet and an older one will reveal a safety gap that cannot be fairly attributed to the software.
Philip Koopman, a professor of electrical and computer engineering at Carnegie Mellon University and one of the field’s most cited autonomous-driving safety researchers, reviewed Tesla’s methodology and was unambiguous. “It’s like saying: ‘My jet airplane is faster than your World War II bomber,'” he wrote. “So, what’s your point?”
Perhaps more significant is what Reuters found among Tesla’s own workforce. Nine former Tesla data labelers (employees who are responsible for training the FSD artificial intelligence by reviewing footage from vehicle cameras) were interviewed for the investigation. Seven said they would not trust FSD to drive them. One described the safety statistics as “bullshit.” Another said he would not ride in a Tesla robotaxi “if you fucking paid me.”
Scandinavian countries remain skeptical. Swedish investigators expressed surprise that FSD routinely exceeds posted speed limits, a behavior that would be unlawful under European traffic law.
Too American
Regulators in Germany and France are joining Sweden, and each has raised the same underlying concern: Tesla’s safety data is derived from American roads, American driving conditions, and American traffic laws.
Its applicability to the European context has not been independently established. Dutch testing flagged that FSD does not reliably detect motorcyclists – a significantly more consequential gap on European roads than American ones.
Norway’s Public Roads Administration, responding to Tesla owners who cited the company’s safety report in lobbying letters, observed that the statistics in question are “self-produced,” making it “difficult to find correlation with the authorities’ accident statistics.”
The Dutch RDW, which was first in Europe to approve FSD, performed a test comprising 1.8 kilometers before positive evaluation. Even though the Flemish infrastructure landscape is very different from that of the Netherlands, due to ribbon development, Minister of Mobility Annick De Ridder seems to rely on those numbers for her own approval. She demanded no more than 5,000 kilometers with only one vehicle.
Currently, the liability of using FSD lies with the driver, who is responsible at all times. The approval is temporary. For authorization to pass fully, member states representing 55 percent of EU countries and 65 percent of the bloc’s population must still vote in favor.
A committee meeting scheduled for the end of the month debated the use of FSD but did not include a vote on its agenda. The next windows are July and October.
