This may be true, but paradoxically, the data might be slightly skewed since Tesla's driver monitoring system (which is only active during autopilot) actively rewards drivers who are paying attention.

So is the accident rate lower because people are forced to be more attentive during FSD? Or is it genuinely lower (i.e, if you took out the driver, would there be less accidents)? To be fair, I'd still wager that yes, FSD is probably statistically way better than the average driver.

Maybe some combination of miles per intervention + accident data would give more insights into that.