Tesla, the US electric vehicle maker developing its Full Self-Driving Supervised driver-assistance system for European roads, faces new scrutiny after Belgian testing found repeated speeding and attempts to overtake cyclists where local traffic rules prohibit the manoeuvre.
Belgian road-safety organisation Johanna tested Tesla FSD Supervised for three days in July, covering roughly 400 km across Brussels and other parts of Belgium.
The group focused on low-speed environments used heavily by pedestrians and cyclists, including 30 km/h zones, residential streets and bicycle streets.
Its study found FSD exceeded the legal limit on 55% of tested 30 km/h road segments in the Brussels Capital Region. During those speeding instances, the vehicle travelled at an average route speed of 44 km/h, with higher averages recorded when surrounding traffic left the road clear.
Johanna said the vehicle frequently showed the supervising driver an incorrect speed limit even after its cameras appeared to recognise the correct roadside sign.
In numerous 30 km/h zones, FSD initially recognised a 30 km/h sign but later displayed 50 km/h as the applicable limit. The organisation said incorrect information on the screen risks misleading drivers responsible for supervising the system and appears connected with some of the recorded speeding.
Testing in Belgian residential areas produced similar results at lower legal limits. FSD exceeded the 20 km/h limit at 42% of the locations examined by Johanna.
Bicycle streets produced another compliance problem. Belgian traffic rules prohibit motor vehicles from overtaking cyclists on these roads, yet videos recorded during the study show FSD repeatedly attempting overtaking manoeuvres.
Johanna said the system appeared unable to distinguish the Belgian rule from road rules in neighbouring countries, including the Netherlands, where requirements differ. The finding raises questions about how an AI-based driving system interprets country-specific traffic regulations once deployed across several European markets.
The organisation also documented stronger aspects of FSD’s behaviour. During its testing, the system generally approached pedestrians and cyclists cautiously and frequently yielded before vulnerable road users entered its path.
Johanna concluded those results didn’t compensate for the repeated speed-limit violations and bicycle-street errors recorded during the same test programme. The group submitted its findings on September 7 to Dutch vehicle regulator RDW and the Flemish transport ministry, asking both authorities to reconsider their approvals.
RDW granted Tesla FSD Supervised provisional European type approval in April after more than 18 months of testing on closed facilities and public roads.
The regulator describes FSD as a driver-controlled assistance system rather than autonomous driving technology, meaning the person behind the wheel remains responsible for the vehicle at all times.
Belgium subsequently permitted the system, alongside several other European countries. A wider European decision on approval is currently scheduled for October 6, giving Johanna’s report added significance before regulators consider broader deployment.
Several European countries have already raised questions about whether FSD consistently follows local speed restrictions. Sweden is among governments that have expressed concerns, while authorities in countries allowing the system stress that legal responsibility remains with the supervising driver.
Xavier Lesenne, spokesperson for the Flemish government’s transport ministry, described FSD as a learning system processing additional information over time. He said drivers remain fully responsible and must intervene when the software makes an error or fails to apply a traffic rule correctly.
The Flemish government also pointed to Tesla data suggesting FSD has accident-prevention potential. Tesla has submitted extensive European testing material to regulators as it seeks broader approval for the software.
Tesla’s own evidence presents a different picture of FSD’s speed behaviour. In an analysis of 708 randomly selected European driving samples where the system set a maximum speed above its detected speed limit, Tesla reported that FSD travelled at or below the median speed of surrounding traffic in 98% of higher-speed cases.
Tesla said none of the vehicles in the remaining high-speed samples travelled faster than the fastest surrounding vehicle. The company argues its contextual speed system adjusts to traffic flow rather than producing excessive speeding, including situations where detected speed-limit information is inaccurate.
Those tests differ materially from Johanna’s work. Tesla’s analysis examines how FSD compares with surrounding traffic after increasing its maximum speed, whereas the Belgian study focuses on compliance with posted legal limits in low-speed urban environments.
Tesla also reports almost 1 mn miles of European FSD Supervised validation driving across eight countries, with no major or minor collisions attributed to system performance. Its fixed-route testing covered more than 230,000 predefined road interactions across six European cities.
FSD also needs to operate within national traffic laws, including rules governing speed and interaction with cyclists.
The European Transport Safety Council has previously raised concerns about applying one automated-driving system across countries with different road regulations. Belgium’s bicycle-street rules offer a practical example of the problem, since a manoeuvre permitted in one jurisdiction isn’t necessarily legal in another.
RDW said after granting provisional approval that its evaluation included European driving data and independent statistical review of Tesla’s evidence. The regulator continues monitoring FSD performance on public roads as more data becomes available.
Johanna wants the Dutch regulator and Flemish authorities to incorporate its July road tests into that assessment before wider European approval proceeds.
The group published the report on September 24 to give regulators and other parties time to examine its findings ahead of the planned October vote.
