Key numbers
We rely on official statistics, Tesla's quarterly Vehicle Safety Reports (2025 data), Tesla's European evidence dashboard (2026), fleet estimates, and peer-reviewed insights on autonomous vehicle (AV) safety.
Evidence base
Tesla's 2025 Vehicle Safety Reports show that when FSD (Supervised) is engaged:
This equates to an ~82–86% reduction in crashes compared to average human driving, depending on severity. Using a 6× midpoint yields an approximate 83% reduction.
Tesla's November 2025 FSD safety data
| Metric | FSD (Supervised) | U.S. National Average | Safety Factor | Approx. Reduction |
|---|---|---|---|---|
| Miles per minor collision | ~986,000 | ~178,000 | ~5.5× | ~82% |
| Miles per major collision | ~5,000,000 | ~699,000 | ~7× | ~86% |
This is substantial progress (and better than non-FSD Tesla driving), but not the ~10× factor needed for a 90% reduction. Autopilot (mostly highway use) performs better (~6–7 million miles per crash, closer to 9–10× the average), but FSD operates in complex urban environments, so its gains are lower.
The 94% "Human Error" Statistic — Context
A commonly cited claim is that FSD could achieve a 90% crash reduction because human error causes 94% of crashes. This originates from NHTSA's 2008 study (summarized in 2015) analyzing ~2 million crashes, where researchers assigned the "critical reason" to the driver in ~94% of cases.
However, this statistic is often misused when applied to self-driving technology:
- The study only examined human-driven vehicles, so it does not mean 94% of crashes are solely caused by humans or would be preventable by automation
- Many crashes involve multiple factors (road design, weather, other drivers) — the "critical reason" is not full causation
- The study doesn't account for new error modes that autonomous systems introduce
What FSD Does and Doesn't Eliminate
- Distraction and impairment — Yes, FSD avoids these (no texting, fatigue, or DUI)
- Decision errors — Partially: the AI can still make poor judgments, such as misinterpreting situations, hesitating, or taking unsafe actions. Tesla faces ongoing NHTSA probes into FSD-related crashes
- Novel failure modes — AVs introduce sensor failures in bad weather, software bugs, and inability to handle edge cases (unpredictable pedestrians, construction zones)
Tesla's data is self-reported and has faced methodology criticism (e.g., selection bias in when features are engaged, definitions of "crash"). Independent verification is limited. Some analysts highlight a "human driver paradox": the safer FSD becomes, the more drivers may disengage attention, potentially offsetting gains.
Tracking start date
Our live counters begin tracking from November 21, 2025 — the date Tesla began the limited rollout of Full Self-Driving (Supervised) v14.2 (part of firmware 2025.38.9.5).
Countries approved after that date accrue "lives saved" only from their own approval date, and accrue "lives lost to regulation" from November 21, 2025 until that date (frozen thereafter): Netherlands April 10, 2026; Lithuania May 20; Estonia May 29; Denmark June 9; Belgium June 11; Slovenia September 8.
This version, primarily targeting Hardware 4 (HW4) vehicles, featured:
- Upgraded neural network vision encoder
- Improved object detection capabilities
- Updated, more detailed UI visualizations
This release marked a significant step toward supervised autonomy at scale, making it a logical baseline for tracking impact.
83% for current supervised FSD estimates (6× safety factor midpoint, based on Tesla's November 2025 FSD-specific safety report). For serious safety outcomes lean toward 86% (7×); for broader incident reduction, 82% (5.5×) is more cautious.
European cross-check
Tesla's FSD Evidence Dashboard, September 2026.
On September 1, 2026 Tesla published its Article 39 FSD (Supervised) Evidence Dashboard, the safety file it submitted to European regulators ahead of the EU-wide vote on October 6. It is the first FSD safety dataset measured on European roads, so we use it to sanity-check the U.S.-derived reduction factor above.
| Metric (EU, HW4 vehicles) | Tesla Evidence Dashboard | This tracker |
|---|---|---|
| Countries with FSD Supervised | 5 at publication — Netherlands, Lithuania, Estonia, Denmark, Belgium | 6, adding Slovenia (Sep 8), with per-country approval dates |
| Active users | 70,000+ | 71,000 (split by fleet size: NL 39k, DK 16k, BE 13k, EE 1k, LT 1k, SI 1k) |
| Distance driven | 100M+ km since April 10; ~1M km per day | Not modelled directly (market-share formula) |
| Crash likelihood vs manual driving | 4.1× lower overall (~76% reduction); 5.2× in the April–July interim set (~81%) | 6× (83%) — from U.S. November 2025 data |
| Major crashes by road type | −45% highway, −69% arterial, −50% city, −40% local | Single blended 83% |
| Other indicators | Up to 84% fewer automatic emergency braking events; ~66% lower driver fatigue; zero collisions over 16.6M highway km in the Netherlands | — |
| Pre-approval validation | 793,000+ test miles in 8 countries, 230,000+ scenario tests in 6 cities, 18+ months RDW evaluation | — |
Fleet size, user counts and country list match. The reduction factor does not: Tesla's own European figure is 4.1× (76%) against the 6× (83%) we apply everywhere. Applying 76% instead of 83% to the five EU countries would lower their "lives saved" estimate by about 9%. Their combined contribution is under 6 lives per year, so the global counters move by well under 1%. We keep the single 83% factor for consistency and note that for Europe it is the optimistic end of Tesla's own data. The 4.1× figure measures all crashes; the per-road-type numbers measure major crashes only, and Tesla has not published the collision definitions or the manual-driving baseline behind them.
Current impact
FSD (Supervised), September 2026.
FSD (Supervised) is active on ~1.48 million vehicles globally (Tesla Q2 2026: up from 1.1M in Q4 2025 and 1.28M in Q1 2026), skewed toward the U.S. The per-country user figures in the tracker still use the Q4 2025 split for North America and Tesla Europe's September 2026 figure for the EU.
Key adjustments
- Only ~20–30% of miles by active users are typically driven with FSD engaged (conservative estimate based on usage patterns and cumulative data trends)
- Effective FSD-engaged miles: ~4–6 billion annually (derived from fleet size, adoption, and growth toward 10+ billion cumulative miles in 2026)
- Reduction: 83% (6× safety factor from November 2025 FSD-specific data)
Miles-based formula
Example Calculation (U.S.-focused, ~90% of impact)
| Metric | Value | Source |
|---|---|---|
| U.S. fatality rate | ~1.26 deaths per 100M VMT | NHTSA 2023–2025 |
| Engaged miles (U.S. dominant) | ~4–5 billion/year | Estimated |
| Expected fatalities if manually driven | ~50–65 | Calculated |
| With FSD engaged (83% reduction) | ~40–54 lives prevented/year | Calculated |
This number is small but growing exponentially with adoption, software improvements, and new rollouts. It represents real, observed safety gains today.
What we track
🟢 Lives Saved (FSD Allowed)
Regions where FSD is legally permitted and active. We count actual safety improvements based on fleet size and engagement rates.
🔴 Lives Lost to Regulation
Regions where Teslas exist but FSD (Supervised) is blocked. We estimate the lives that could be saved if supervised deployment were permitted at observed engagement rates.
The "Lives Lost to Regulation" counter uses the same supervised impact assumptions (engagement rate and 83% reduction) and applies them to Tesla fleets in regions where supervised FSD is not permitted.
Limitations
- Engagement rate assumptions — Current estimates depend on engagement rate assumptions and assume crash reductions translate proportionally to fatalities
- Fleet data is estimated — Tesla does not break out per-country data; estimates from third-party tracking
- Network effects not modeled — Does not account for network effects (safety improves faster at higher penetration) or indirect benefits (e.g., smoother traffic flow)
- At-fault assumption simplified — Real-world crash responsibility is complex and varies by scenario
Data tables
Below are the datasets we use for calculations. All data is sourced from official statistics and third-party fleet tracking as of February 2026.
Tesla Fleet & FSD Users by Country
Tesla disclosed 1.1 million active FSD users/subscribers globally in Q4 2025 earnings (January 2026). ~70% purchased outright, ~30% subscription. Global fleet ~9 million vehicles as of early 2026. Estimates based on delivery trends, regional reports, and third-party registration data.
| Country/Region | FSD Status | Tesla Fleet | Est. FSD Users |
|---|---|---|---|
| 🇺🇸 United States | Active | 4,800,000 | ~950,000 |
| 🇨🇳 China | Active | 2,400,000 | ~45,000 |
| 🇨🇦 Canada | Active | 325,000 | ~90,000 |
| 🇦🇺 Australia | Active | 175,000 | ~7,000 |
| 🇲🇽 Mexico | Active | 125,000 | ~8,000 |
| 🇰🇷 South Korea | Active | 100,000 | ~3,000 |
| 🇳🇿 New Zealand | Active | 35,000 | ~2,000 |
| 🇩🇪 Germany | Blocked | 375,000 | — |
| 🇬🇧 United Kingdom | Blocked | 275,000 | — |
| 🇳🇴 Norway | Blocked | 225,000 | — |
| 🇫🇷 France | Blocked | 175,000 | — |
| 🇳🇱 Netherlands | Active | 135,000 | ~39,000 |
| 🇸🇪 Sweden | Blocked | 90,000 | — |
| 🇩🇰🇧🇪🇪🇪🇱🇹🇸🇮 DK, BE, EE, LT, SI | Active | 112,000 | ~32,000 |
| Other Europe | Blocked | ~385,000 | — |
| Global Total | ~9M | ~1.1M |
Road Fatalities by Country/Region
Latest official statistics from WHO, NHTSA, European Commission, and national sources.
| Country/Region | FSD Status | Annual Deaths | Rate/100K |
|---|---|---|---|
| 🇺🇸 United States | Active | ~39,000 | 11.5 |
| 🇨🇳 China | Active | ~60,000 | 4.3 |
| 🇨🇦 Canada | Active | ~2,000 | 5.0 |
| 🇦🇺 Australia | Active | ~1,200 | 4.5 |
| 🇲🇽 Mexico | Active | ~16,000 | 12.0 |
| 🇰🇷 South Korea | Active | ~3,000 | 5.8 |
| 🇪🇺 EU (20 states still blocking) | Blocked | ~18,500 | 4.5 |
| 🇬🇧 United Kingdom | Blocked | ~1,600 | 2.4 |
| Global Total | ~1.19M | ~15.0 |
Impact Calculation by Country
Formula: Lives/Year = Annual Deaths × 0.83 × (FSD Users or Fleet ÷ Total Vehicles)
| Country | Calculation | Result |
|---|---|---|
| 🇺🇸 United States | 39,100 × 0.83 × (950,000 ÷ 136,000,000) | ~227/yr |
| 🇨🇳 China | 60,716 × 0.83 × (45,000 ÷ 564,800,000) | ~4/yr |
| 🇨🇦 Canada | 2,000 × 0.83 × (90,000 ÷ 16,000,000) | ~9/yr |
| 🇦🇺 Australia | 1,170 × 0.83 × (7,000 ÷ 10,400,000) | ~0.7/yr |
| 🇲🇽 Mexico | 15,600 × 0.83 × (8,000 ÷ 52,000,000) | ~2/yr |
| 🇰🇷 South Korea | 3,016 × 0.83 × (3,000 ÷ 20,800,000) | ~0.4/yr |
| 🇳🇿 New Zealand | 286 × 0.83 × (2,000 ÷ 2,040,000) | ~0.2/yr |
| Total FSD Active | ~243/yr | |
| 🇩🇪 Germany | 3,078 × 0.83 × (375,000 ÷ 33,280,000) | ~29/yr lost |
| 🇬🇧 United Kingdom | 1,615 × 0.83 × (275,000 ÷ 26,920,000) | ~14/yr lost |
| 🇳🇴 Norway | 116 × 0.83 × (225,000 ÷ 2,200,000) | ~10/yr lost |
| 🇫🇷 France | 3,187 × 0.83 × (175,000 ÷ 27,120,000) | ~17/yr lost |
| 🇳🇱 Netherlands | 669 × 0.83 × (39,000 ÷ 7,040,000) | ~3/yr |
| 🇸🇪 Sweden | 210 × 0.83 × (90,000 ÷ 4,200,000) | ~4/yr lost |
| Other Europe | ~10,500 × 0.83 × (~385,000 ÷ ~94,000,000) | ~36/yr lost |
| Total FSD Blocked | ~110/yr lost |
Why this matters
Road traffic deaths (~1.19 million globally) remain one of the leading preventable causes of death. Tesla's data-driven approach demonstrates real progress, with current FSD already delivering measurable safety gains.
This tracker highlights both today's impact and the urgency of evidence-based regulation to unlock greater benefits.
Primary Sources
Further improvements welcome—transparency drives better outcomes. If you have better data sources or methodology suggestions, contact us at hello@giagolab.com