A Reuters investigation has put Tesla's Full Self-Driving safety claims in the spotlight at the worst possible moment. Using public records requests, the news agency obtained government communications showing that Tesla submitted self-compiled safety statistics to road traffic regulators in Sweden and the Netherlands as part of its campaign to secure FSD approval across Europe. The European Transport Safety Council (ETSC) and independent road safety researchers have since reviewed the data — and their verdict is damning: the statistics are structured in a way that makes FSD look far safer than any honest comparison would support.
The Numbers at the Center of the Dispute
Tesla has built its regulatory case around two headline figures:
- FSD-equipped vehicles are up to 10× safer than human drivers
- The average crash-free interval for FSD vehicles is 7× longer than the U.S. fleet average
These are not invented numbers. They are drawn from Tesla's own vast dataset. The dispute is not about whether the data exists — it is about whether the comparisons are constructed fairly.
Swedish transport authority investigator Anders Eriksson put it plainly: "Swedish regulators do not just look at the headline of a report. Any assessment of such an automated driving system will absolutely not rely solely on safety claims packaged by the company itself, but on comprehensive evidence."
ETSC and other European researchers went further, calling the submitted statistics "misleading marketing" and "unreliable pseudo-data."
Three Statistical Problems That Undermine the Claims
1. Apples vs. Oranges: Mismatched Crash Definitions
Tesla's safety reports count only crashes severe enough to trigger airbag deployment for the FSD side of the comparison. On the other side, they use the U.S. national crash rate — which includes every reported incident, from major collisions to minor parking-lot scrapes and low-speed fender-benders that would never deploy an airbag.
The math is straightforward: if you filter your own data to include only the worst crashes while measuring your competitor using all crashes, your numbers will look dramatically better. A legitimate comparison would apply the same severity threshold to both datasets.
2. New Cars vs. Old Cars — Not FSD vs. No FSD
The U.S. fleet average that Tesla uses as its baseline is exactly that: an average of everything on American roads. That includes aging pickup trucks, motorcycles, commercial freight vehicles, and passenger cars manufactured well before automatic emergency braking, forward collision warning, and modern structural safety standards were commonplace. The average U.S. vehicle is approximately 12 years old.
Tesla's FSD vehicles, by definition, are new cars equipped with the most advanced passive safety technology available. Comparing a brand-new Tesla against a fleet average that includes 2003-era trucks does not measure the safety benefit of FSD. It measures the safety benefit of driving a new car — and then attributes the entire gap to the software.
3. Who Actually Uses FSD?
The drivers who opt into FSD are not a random cross-section of American motorists. They tend to be more technologically engaged, more aware of the system's capabilities and limitations, and more likely to activate FSD in the structured, well-mapped conditions where it performs reliably. This self-selection effect means that even without FSD turned on, this group might have a lower crash rate than the national average. The data cannot distinguish between "FSD makes driving safer" and "cautious, tech-savvy drivers who use FSD are already safer drivers."
Why the Timing Matters
This controversy has landed at a particularly sensitive moment in Tesla's European regulatory push. The Netherlands' RDW — the lead European authority evaluating FSD — granted approval for Dutch testing earlier this year. Sweden separately authorized supervised FSD testing on public roads. Tesla has been expanding FSD demonstrations across the continent in anticipation of the prize it has been working toward: a pan-European cross-border approval vote that would allow FSD to operate across EU member states under a single unified framework.
The Reuters investigation, and the ETSC's public response, has introduced a complication that European regulators cannot easily set aside. If the safety data submitted to support approval is methodologically flawed, the approval process must either demand better evidence or move forward on a weaker foundation. Neither outcome is comfortable for Tesla or for the regulators themselves.
What the Data Does — and Doesn't — Tell Us
It is important not to collapse two separate questions into one:
Does FSD make driving safer? Probably yes, under the conditions it is designed to handle, and for the population of drivers who use it. Tesla's FSD fleet has now accumulated over 8 billion supervised miles — a dataset that, methodological caveats aside, contains a genuine safety signal. No credible analyst argues the technology has zero value.
Is FSD specifically 10× safer than the average American driver? This claim, as constructed, does not survive scrutiny. The 10× figure depends on comparison choices — severity filtering, fleet composition, user self-selection — that all systematically tilt the result in Tesla's favor. Change any one of those choices to a more neutral standard, and the multiplier shrinks considerably.
For everyday consumers, this distinction may seem academic. For European regulators, it is the entire question. They are not being asked whether FSD has some safety value. They are being asked whether it meets specific, quantifiable safety thresholds required for public road operation across dozens of countries. For that determination, methodology is not a footnote — it is the foundation.
What Happens Next
The Reuters investigation does not stop the European approval process. It complicates it. Several paths forward are plausible:
Most likely: Regulators request independent third-party crash data using consistent severity definitions, and a comparison against a matched vehicle cohort — same age, same passive safety equipment — rather than the full U.S. fleet. This delays the pan-European vote by months.
Possible: Tesla submits a revised methodology with more conservative comparison standards. The revised numbers would almost certainly show a smaller safety advantage — but a credible one, which may ultimately serve Tesla's long-term regulatory interests better than inflated figures that invite this kind of scrutiny.
Less likely: Individual countries proceed independently on their own timelines, creating a fragmented patchwork of approvals while the pan-European vote stalls.
The Bigger Picture
The core issue here is not whether Tesla's FSD is dangerous. The evidence suggests it is not. The issue is whether the company chose to present its safety data to regulators in the most flattering possible light rather than the most accurate one — and whether European regulators, now aware of the methodological problems, will accept that as sufficient grounds for approval.
The ETSC's characterization of the data as "misleading marketing" is pointed language for a technical safety body to use in an official context. It signals that the evidentiary bar for FSD's European approval has just gotten higher, regardless of what the underlying data might eventually show when measured fairly.
Tesla may well have built a genuinely safer driving system. Proving it, however, requires numbers that can withstand independent scrutiny — not numbers that are optimized to impress.