Tesla Releases FSD Crash Data That Appears More Honest (www.forbes.com)

Tesla published new crash statistics for its supervised Full Self-Driving (FSD) system using miles-driven-per-collision metrics, replacing its criticized Autopilot-only reports. The company now separates highway vs. non-highway and North America vs. global results, and categorizes “major” collisions (airbag or pretensioner deployment) vs. minor ones. Tesla reports that FSD-equipped cars supervised by drivers show ~1.5x fewer collisions on city streets globally compared with similar Teslas, about 2x in North America, and roughly 4x fewer collisions compared with much older Teslas lacking ADAS. This level of granularity makes direct “apples-to-apples” comparisons easier than the prior Autopilot reports, which largely compared freeway airbag-triggering crashes to general police-reported crashes and produced misleadingly large safety claims. Significant caveats remain: Tesla’s non-Tesla “U.S. Average” uses towed vehicles as a proxy for major crashes (not directly comparable to airbag data), the company hasn’t released historical quarters for trend analysis, and the dataset includes all crashes, not only at-fault incidents. Human supervision effects (a helpful “second set of eyes” versus driver complacency or misuse) likely produce mixed real-world outcomes. The data tentatively supports that supervised FSD improves safety on balance, but independent scrutiny and quarterly trendlines are needed—especially because Tesla’s stated robotaxi goal requires FSD alone to outperform humans by a much larger margin (2–3x) before unsupervised deployment would be credible.