Tesla’s quarterly Vehicle Safety Report is Tesla’s own miles-and-crashes table for Autopilot / FSD vs a baseline Tesla constructs. The Q2 2025 recap Tesery originally ran treated FSD as a “game-changer.” The document is real; the causal claim is Tesla’s, not NHTSA’s.
Key facts
- Source: Tesla’s public Vehicle Safety Report — company definitions of “miles,” “crash,” and which driver-assist mode was on.
- Sister numbers that quarter: Tesla said it produced 410,244 vehicles and delivered 384,122 in Q2 2025.
- What the report cannot do: Prove FSD is safer than a matched human on the same roads. Selection bias (who uses FSD, where) is the standard critique.
How to read Tesla’s crash ratios
Tesla compares crashes per million miles with Autopilot/FSD on vs off, and sometimes vs NHTSA all-roads averages. Those denominators are not the same mix of highway vs city. Regulators have separately pressed Tesla on FSD data — see Tesery’s NHTSA data-audit piece.
FSD remains Supervised on owner cars. A safety PDF is not a driverless license.
FAQ
Did NHTSA write this report?
No. Tesla did.
Does a better quarter mean FSD is solved?
No. One quarter of Tesla-defined miles is not a closed safety case.
If you use Autopilot or FSD, you still supervise. The report is Tesla’s scoreboard.
