What exactly Tesla claimed — and why it doesn't add up
According to correspondence obtained by Reuters through public document requests, Tesla approached the Dutch road authority RDW in November 2024 seeking FSD approval. In the letter, it referenced its safety report and claimed that "increased use of FSD leads to safer roads."
After more than a year of testing, the RDW approved FSD in the Netherlands in April 2026 and is now seeking EU-wide approval. Shortly thereafter, Tesla policy manager Ivan Komusanac wrote an email to Swedish regulators with a presentation containing claims that Tesla vehicles with FSD travel more than seven times the distance between accidents than the average human driver in the US. The presentation also claimed that FSD could have saved 32,000 lives and prevented 1.9 million injuries.
However, independent researchers contacted by Reuters labeled these figures as highly misleading. They identified several problems:
- Unrealistic assumption: Calculations are based on a hypothetical scenario where every vehicle in the US — including trucks and accident-prone motorcycles — would be replaced by a Tesla with active FSD.
- Incomparable accident categories: Tesla compares the number of accidents where airbags deployed in FSD-equipped vehicles with the overall accident rate of all vehicles in the US, which also includes far less serious accidents.
- Unfair vehicle age comparison: New Teslas are compared to the average American vehicle, which is significantly older and lacks modern safety features.
Regulator and OECD reactions: "Unreliable data"
The OECD has added the case to its AI Incidents and Hazards Monitor (AIM) — a global database that tracks incidents and potential threats associated with artificial intelligence. The incident is classified as an "AI hazard" (potential threat), not as an already-occurred incident — precisely because no accident has yet occurred based on the misleading data, but the risk is real.
According to the OECD, the event violates the principles of transparency and accountability in deploying AI systems. The affected stakeholders are primarily consumers and the general public.
Dudley Curtis, spokesperson for the European Transport Safety Council (ETSC), told Reuters that the organization is "definitely concerned" that Tesla submitted "unreliable safety data" to Swedish authorities. He added: "If Tesla wants to make safety claims, it should give the data to a university, have it independently verified by a qualified researcher, and then let's talk."
Swedish regulator Anders Eriksson of Transportstyrelsen stated that they "don't just look at headlines" and that any assessment of the system would not be based "solely on aggregated safety claims, but on the overall evidence presented."
What this means for European drivers
The FSD system — despite its name — is not fully autonomous. It requires constant driver attention, who must be ready to take over driving immediately. In the EU, it is classified as an "assistance" system, not autonomous. Tesla charges a monthly subscription for FSD, and Elon Musk has called its approval in Europe crucial for sales growth on the continent, where the automaker lost significant market share last year.
For EU-wide FSD approval, representatives of 55% of member states representing 65% of the EU population must vote in favor. Until that happens, individual states can approve the system independently — the Netherlands has already done so, and Greece has announced the same intention.
The Norwegian road authority responded laconically to letters from Tesla fans who cited company safety statistics and demanded expedited approval: "Tesla's data is self-produced, which makes it difficult to correlate with official accident statistics."
Czech context: why it matters here too
The Czech Republic, as an EU member state, will be one of those voting on any prospective EU-wide FSD approval. Thousands of Tesla vehicles equipped with FSD hardware already drive on Czech roads — but the system currently operates in a limited mode because European certification is lacking.
Furthermore, under the AI Act, the EU classifies autonomous driving as a "high-risk" area of artificial intelligence, meaning stricter requirements for transparency, testing, and safety documentation. Tesla's case shows that without independent auditing and standardized testing protocols, there is a risk that regulators will approve systems based on marketing materials rather than verified reality.
Broader context: Trust in AI behind the wheel
The case comes at a time when public trust in autonomous vehicles is, according to surveys, rather declining. Every incident, even just a potential one, further undermines people's willingness to entrust driving to artificial intelligence. The OECD AIM monitor itself records dozens of incidents in the "mobility and autonomous vehicles" category — from erroneous AI decisions to accidents to regulatory disputes like this one.
For Tesla, this is yet another in a series of controversies surrounding FSD. In the US, it faces an NHTSA investigation (National Highway Traffic Safety Administration) over accidents linked to Autopilot and FSD. The European case, however, adds a new dimension — the question of whether technology companies can self-certify as safe without independent oversight.
Is Tesla's FSD truly fully autonomous driving?
No. Despite the name "Full Self-Driving," it is an SAE Level 2 assistance system that requires constant driver supervision, with the driver ready to intervene immediately. Fully autonomous driving (SAE Level 4–5), where the driver does not need to monitor traffic at all, is not yet commercially offered by any automaker.
What is the OECD AIM and why is its classification important?
The OECD AI Incidents and Hazards Monitor (AIM) is a global database that automatically monitors and documents incidents and potential threats associated with artificial intelligence. It serves as a foundation for regulation and policy-making. The case's inclusion in AIM means that an international organization considers the situation worthy of regulatory attention — it is not a legal decision, but a significant signal to lawmakers.
When will FSD be available in the Czech Republic?
That depends on two factors: EU-level approval (requiring consent from 55% of member states representing 65% of the population) and subsequent implementation into national legislation. Given the current controversy and the AI Act's requirements for high-risk AI systems, the approval process is likely to take at least several months to years. Individual states — so far the Netherlands — can, however, approve the system independently.
Sources: OECD AI Incidents Monitor, Reuters / Livemint, Electrek
