Tesla’s European FSD Safety Data: What the 4.1× Claim Actually Shows

Tesla’s European safety case for Full Self-Driving (Supervised) rests on a company-produced comparison of real-world fleet collision rates. Tesla said vehicles using the system had 4.1 times fewer collisions per kilometre than manually driven Teslas in the five European countries where the feature had provisional permission, based on about 102 million km of FSD-supervised driving between April and August 2026. 12

That is a notable claim, but it should not be read as an independent determination that FSD causes a 4.1-fold safety improvement. The data, definitions, exposure controls and underlying regulatory testing remain central to assessing the result.

Tesla’s headline numbers

Tesla’s comparison covered the Netherlands, Denmark, Belgium, Estonia and Lithuania. The company said FSD (Supervised) was used by more than 70,000 customers in those markets, with the fleet travelling more than 1 million km per day. 215

The published collision counts were:

Road type	FSD (Supervised)	Manually driven Teslas
Highways	3 collisions	137 collisions
Non-highway roads	9 collisions	490 collisions

Tesla’s headline comparison was based on collision rates per kilometre rather than these raw totals, because the FSD and manual-driving groups accumulated different distances. 7

The key takeaway is therefore not that 3 is directly comparable with 137, or 9 with 490. Tesla’s reported 4.1× result is a distance-adjusted aggregate across the five countries and road environments. 12

What Tesla put in its dashboard

On 1 September 2026, Tesla published an open-source FSD evidence dashboard that it said had previously been provided to regulators. The dashboard assembled nine findings across three datasets, including fleet collision comparisons, driving-performance indicators, lane-change analysis, engineering-fleet validation and fixed-route testing in European cities. 2315

Tesla framed publication as a transparency measure and as evidence for its request to expand supervised FSD beyond the countries that had already granted permission. 115

Tesla’s own dashboard also reports a range of collision-rate reductions across road classes when comparing FSD (Supervised) with manually driven Tesla vehicles that had active safety features enabled. Those figures are model-based comparisons, expressed as incident-rate ratios, rather than an independent regulator audit. 13

Which countries permitted FSD (Supervised)?

The Netherlands was the first country to grant provisional approval in April 2026. Belgium, Denmark, Estonia and Lithuania followed, bringing the total to five. 1415

Tesla was seeking broader EU authorization to deploy supervised FSD across the bloc. Reporting at the time described a possible EU-wide vote on 6 October 2026 and said wider adoption required support from at least 15 of the EU’s 27 member states representing 65% of the EU population. 46

For Tesla, wider permission would make the feature available across a much larger European market rather than being limited to five countries. It would also determine whether the company could more broadly market and deliver one of its prominent software features to European customers.

Why the figures are contested

The main limitation is not that fleet data are useless; it is that observational fleet comparisons are difficult to interpret without detailed, independently reviewable methods.

A per-kilometre comparison can be affected by differences in the road types, traffic conditions, weather, time of day, driver behaviour and trip complexity in each group. Drivers may also choose when to activate FSD, which can make FSD and manual trips systematically different before the software’s performance is considered. Tesla’s dashboard notes that fleet analyses cannot fully control for factors such as traffic, weather, lighting and scenario complexity, and that selection effects and residual confounding may remain. 13

Independent traffic-safety researchers told Reuters that self-published Tesla statistics submitted to European regulators amounted to misleading marketing. Reuters also reported that critics challenged earlier Tesla safety claims for comparing crashes in FSD-driven Teslas that triggered airbag deployment with a broader U.S. crash rate that includes less-severe accidents.

That criticism concerns a different comparison from the five-country 4.1× fleet claim, but it explains why outside reviewers have called for clearer definitions, matched exposure data and independent analysis before treating Tesla’s numbers as a settled safety verdict.

Transparency and commercial-confidentiality concerns

Tesla’s public dashboard made more of its selected evidence accessible. However, it did not eliminate disputes over access to the underlying safety case.

Reuters reported that Tesla had pressed the Dutch vehicle authority, RDW, to keep documentation from the FSD safety review confidential, while the regulator said disclosure could violate Tesla’s commercial secrets. European vehicle-safety law experts cited in that reporting argued that too much had been withheld from public scrutiny.

In the United States, two senators also asked the national traffic-safety regulator to examine Tesla’s self-published FSD crash statistics after a Reuters investigation questioned the company’s presentation of safety data.

Bottom line

Tesla’s evidence dashboard presented a large European fleet dataset and a clear corporate claim: 4.1 times fewer collisions per kilometre for FSD (Supervised) than manual Tesla driving, with 3 versus 137 highway collisions and 9 versus 490 non-highway collisions reported before mileage adjustment. 17

Those results may be relevant to regulators and customers, but they are not equivalent to an independently validated causal safety assessment. A robust conclusion would require transparent crash definitions, exposure and road-context data, and review that can test whether the FSD and manual-driving groups are genuinely comparable. 13