Tesla’s Article 39 dashboard leaves France and Sweden less room to object
Tesla has dropped a new FSD Europe data pack on tesla.com. This data was used as part of the EU Article 39 file which was the regulation used to get a temporary type-approval from the RDW. The data includes collision rates, indicator use, fatigue, gaze, speed matching, city routes.
This is part 4 of Tesla FSD Regulations for the UK and Europe. Parts 1 to 3 discuss the regulations for FSD (Supervised) in Europe and the UK, what UN R171 Series 02 actually changes, and when a new UN certificate could be implemented or FSD approval.
1. Why did Tesla release this European FSD data?
Tesla titles it Article 39 - FSD (Supervised) Dashboard. It is meant to support Article 39(2)(b) of EU Regulation 2018/858, used to apply for an exemption, enabling FSD on the roads in the EU. The Dutch RDW issued an approval based on this data on 10 April 2026. Other member states were handed this data too, but still most EU members refuse to adopt the type-approval issued by the RDW. But Tesla have played a 4D chess move here by releasing all the data, showing that many concerns are unjustified.
2. The numbers that actually matter for FSD approval in Europe and the UK
"Study 6" within the document relates to collision data and the numbers speak for themselves: More than 793,000 FSD Supervised miles across eight European countries, and zero major or minor collisions put down to FSD. Anyone saying “there is no European collision evidence” is now misinformed.
The rest is about how the car behaves, which is what the exemptions are for. Tesla lists various requests for exemption against UN R171.01.
- System-initiated manoeuvres
- Withholding of hands-on requests
- Lateral acceleration above the DCAS fixed cap
- Raising the speed cap to match surrounding traffic
France’s urban case and Sweden’s speed-offset case are part of the latter exemption request. Their arguments are seemingly overstated and we can now see the same data they have:
| Metric (Europe, Tesla dashboard, 3 Sep 2026) | FSD (Supervised, HW4) | Manual with active safety |
|---|---|---|
| Blinker events per 1,000 miles, highway | 1,032 | 834 |
| Blinker events per 1,000 miles, non-highway | 2,328 | 1,567 |
| Lateral accel ≥3 m/s² per 1,000 miles, highway | 7.4 | 29.5 |
| Lateral accel ≥3 m/s² per 1,000 miles, non-highway | 304 | 1,087 |
| Highway drowsiness events per 1,000 miles | 0.445 | 1.34 (67% drop, IRR 0.333) |
| Right-side lane-change awareness | 91.8% | 81.4% |
| DMS warning in the 7 seconds before a lane change | 2.5% (North American DMS study) | — |
| High-speed clips at or below median traffic | 98% of 680 | 0 faster than the fastest nearby car in Tesla’s chart; live page says never exceeded fastest in 99.9% of cases |
| Six-city fixed loops, safety-critical fails | 0 | Live Study 7 also: 230,000+ tests, 99%+ ordinary pass |
| Europe-native interaction tests, safety-critical | 100% | Live headline 230,000+ tests across 12 types |
Those are Tesla’s figures from their engineering fleet. And they show how safe Tesla's FSD (Supervised) system is in Europe.
3. City streets with FSD in Europe
Tesla’s 2024 versus 2025 minor-collision chart, (after v13), also has some interesting data: Arterial roads were 42.6% safer (1.013 down to 0.581 per million miles). Urban collectors at 39.8% safer (1.711 to 1.029). Local access was 42.6% safer (2.523 to 1.449).
France’s position, as we reported in France Declines Tesla FSD Supervised Approval Over Safety Data Concerns, was that the system still needed more convincing safety validation, including in dense urban areas. Transport minister Philippe Tabarot told BFMTV/AFP in July 2026 that there had been progress, but the safety trade-offs were “not yet sufficient for authorisation as-is” (Le Figaro / AFP). He cited speeding, attention in cities, and complex manoeuvres: lane changes, intersections, roundabouts.
However this new data drop (in our opinion) already had all the data they needed to make a decision. A year-on-year drop in minor collisions on arterial and local roads is the sort of number they said they needed, but already had.
4. Indicators, and how hard the car turns
On European engineering miles, FSD uses the indicator more than a manual Tesla with active safety: 1,032 blinker events per 1,000 miles on highway against 834, and 2,328 against 1,567 on non-highway roads.
Lateral acceleration data was also well within acceptable limits for the exemption. Highway events at or above 3 m/s²: 7.4 per 1,000 miles on FSD vs 29.5 on manual driving. Off highway: 304 against 1,087. Tesla’s point here is simple enough, FSD handles the DCAS fixed acceleration cap better than a human driver. The regulators still want the cap though, but perhaps this data is compelling enough to drop the cap in a later revision of the UNECE R171 Series 02 regulation?
5. Fatigue and gaze
France also asked whether driver monitoring still works when the car is doing most of the driving. Tesla’s highway fatigue chart shows a 67% drop in fatigue: 0.445 drowsiness events per 1,000 miles with FSD against 1.34 for manual with active safety enabled. Tesla maps this data to Exemption 1: if you withhold hands-on requests on the motorway, the cabin camera still has to catch someone not paying attention. Lower fatigue than the manual drivers should be the baseline to achieve and they exceeded it significantly.
For gaze, on EU engineering cars, 91.8% of FSD drivers made at least one relevant check (mirror, window, screen or rear-view) before a right-side lane-change, against 81.4% in manual driving. Simply put: FSD does not make drivers less attentive.
6. Sweden, and what the speed sample actually shows
Sweden’s TCMV argument, as we set out in Sweden’s TCMV Objection to Tesla FSD Supervised: Why the Speed Offset Concerns Are Flawed, is the driver-selectable margin above a posted limit. Reuters and SVT reported Trafikverket’s 30 April 2026 letter as recommending a vote against EU-wide approval unless that offset is disabled.
In the data "study 8" reviewed 680 high-speed clips of FSD operating in Europe. In Tesla’s data, 98% ran at or below the median of the cars around them, 12 of 680 sat slightly above. FSD never exceeded the fastest nearby car in 99.9% of cases.
France flagged speed as an issue too. Tabarot’s July example was the system running “up to 50% above the limit” but are these concerns still valid?
7. France, Paris loops, and what is still missing
France’s CNRV position, as we reported in France Tesla FSD Awaiting EU Decision, is that the Dutch decision is a provisional, Netherlands-specific Article 39 approval, not a full European type approval, and that France will wait for the TCMV.
Tesla’s six-city test was following a fixed urban route, run on repeat with a safety driver on board. Tesla says FSD was not trained on those routes before the testing. A fail is a takeover. A safety-critical fail is one that could have hurt someone. Our chart counts 611 loops and ZERO safety-critical fails. - Impressive?
The Europe-native interactions chart is the other interesting data set: 12 types of test, including trams, bikes, roundabouts, lights and school zones. Tesla reports 100% on the safety-critical screen for every single test type.
8. What's next for FSD in Europe?
With the public having access to the same data as the regulators, pressure is on them to vote yes at the upcoming TCMV vote (Date to be decided). Countries can no longer hide behind "we don't have enough data" or "the system is unsafe" remarks.
