Only 1 crash for every 7.44 million miles driven using Tesla Autopilot in Q1 2025

[ModelAZ]

NEWS: Tesla has revealed that in Q1 2025, they recorded one crash for every 7.44 million miles driven in which drivers were using Autopilot technology. For drivers who were not using Autopilot technology, Tesla recorded one crash for every 1.51 million miles driven.

By comparison, the most recent data available from NHTSA and FHWA (from 2023) shows that in the US there was an automobile crash approximately every 702,000 miles.

Sources:

Tesla Vehicle Safety Report:
https://www.tesla.com/VehicleSafetyReport

Sponsored

 

[WheresMyCybertruck]

Does that also include accidents where Autopilot disconnected seconds before because of a potential hazard? Or strictly just accidents where Autopilot was active at the moment of impact?

 

[Gaximus]

I assume that includes all crashes, not just at-fault crashes

 

[DrPhyzx]

Meaningless unless one corrects for the bias that Autopilot gets used more often on roadways with lower accidents/mile; e.g. freeways vs. city streets. The only way to do this is to weight the Autopilot-off accident rate to represent the same mix of miles traveled on the same exact roadways, at the same time periods, as Autopilot-on miles.

 

[SCTesla]

I assume that includes all crashes, not just at-fault crashes

Typically it's at fault crashes. Tesla was, at one point, only reporting crashes which the airbag deployed. Obviously we would need to see the data to verify that. It's any version of AP to include FSD, AP, TACC while it's activated or 5 seconds after activation or it was.

 

[HaulingAss]

Does that also include accidents where Autopilot disconnected seconds before because of a potential hazard? Or strictly just accidents where Autopilot was active at the moment of impact?

Yes, it's attributed as an Autopilot accident even if the driver disengages 5 seconds before impact. And don't believe anyone who says it only counts "at fault" crashes. Tesla is not in the business of determining who was at fault.

These numbers show how laughable it is that the Tesla haters complain Tesla is making the road more dangerous.

 

[Zane Edwards]

NEWS: Tesla has revealed that in Q1 2025, they recorded one crash for every 7.44 million miles driven in which drivers were using Autopilot technology. For drivers who were not using Autopilot technology, Tesla recorded one crash for every 1.51 million miles driven.

By comparison, the most recent data available from NHTSA and FHWA (from 2023) shows that in the US there was an automobile crash approximately every 702,000 miles.

Sources:

Tesla Vehicle Safety Report:
https://www.tesla.com/VehicleSafetyReport

IT IS REALLY THAT GOOD, I use it most of the time. I have taken many family, friends, neighbors etc on rides and let them drive both manual and in FSD and everyone is blown away how good it is. (most people didn't realize it has come this far). For us long time users we have experienced the evolution and continuous improvement of FSD. It is far from perfect but it is really, really good and the light at the end of the tunnel is in sight. These numbers can be interpreted in many ways but I think we can all agree it is a real solution and exciting time for autonomous driving.

 

[Gigahorse]

IT IS REALLY THAT GOOD, I use it most of the time. I have taken many family, friends, neighbors etc on rides and let them drive both manual and in FSD and everyone is blown away how good it is. (most people didn't realize it has come this far). For us long time users we have experienced the evolution and continuous improvement of FSD. It is far from perfect but it is really, really good and the light at the end of the tunnel is in sight. These numbers can be interpreted in many ways but I think we can all agree it is a real solution and exciting time for autonomous driving.

FSD (supervised) is amazing technology.
But I feel like I have intervened and saved 3 accidents in the last 2,000 miles or so.
For the CT it seems like letting one driver for 100,000 miles without intervening would guarantee 1+ accidents, more less 1 million.
7million sounds extremely optimistic

 

[L3it3R]

I regularly go 2-3 hours with FSD engaged and zero input from myself - and it works even better in high density situations in the city (DC and Baltimore area primarily) than it does where I live (rural/suburban). I've been using FSD since late 2021 and V13 on HW4 is by and far the best ADAS I've ever used, and I can't imagine not having it on my vehicles (that's why all 3 have FSD)

 

[CyberTexas]

This is Tesla data…and critics have been critical of this. How legit are these numbers???

I love to read this BTW…trying to convince my “extra rib” to see my way.

 

[Mal]

FSD (supervised) is amazing technology.
But I feel like I have intervened and saved 3 accidents in the last 2,000 miles or so.
For the CT it seems like letting one driver for 100,000 miles without intervening would guarantee 1+ accidents, more less 1 million.
7million sounds extremely optimistic

Would those interventions have caused an accident, or were they just awkward situations?

I've found, lately, that when I intervene it's more that I'm not comfortable with what FSD is doing than it would have caused an accident. For example:

• not leaving enough space after an oncoming car when turning left across oncoming traffic (it's been better lately, and I don't think it would actually have hit anything),
• not moving over one lane when it's in the right lane on a highway and there are cars entering the highway on the right (it does the zipper merge well, but it feels impolite),
• not moving over far enough left when there are, say, folks walking dogs on the right side of the road

These are all sort of human consideration things, which it definitely needs help with. But, for me, I don't think I've had it try to kill me in a pretty long while. I have had it attempt to go straight in a right turn lane recently, but the road was pretty poorly marked and the lane only converted to right turn in the last hundred feet or so. Grandma would have screwed that up, too.

 

[Beetlebug62]

Does that also include accidents where Autopilot disconnected seconds before because of a potential hazard? Or strictly just accidents where Autopilot was active at the moment of impact?

From the Methodology note:

 

[SCTesla]

From the Methodology note:

This is what I was looking for.

They alert on airbag or other active restraint deployed.

A simple rear-end crash wouldn't deploy that, but a Tesla rear-ended extremely hard or into another car would be. There was something on TMC that's how they parse out data.

Obviously things like curb rash, slight sideswipes are not included in the data.

 

[Gigahorse]

Would those interventions have caused an accident, or were they just awkward situations?

I've found, lately, that when I intervene it's more that I'm not comfortable with what FSD is doing than it would have caused an accident. For example:

• not leaving enough space after an oncoming car when turning left across oncoming traffic (it's been better lately, and I don't think it would actually have hit anything),
• not moving over one lane when it's in the right lane on a highway and there are cars entering the highway on the right (it does the zipper merge well, but it feels impolite),
• not moving over far enough left when there are, say, folks walking dogs on the right side of the road

These are all sort of human consideration things, which it definitely needs help with. But, for me, I don't think I've had it try to kill me in a pretty long while. I have had it attempt to go straight in a right turn lane recently, but the road was pretty poorly marked and the lane only converted to right turn in the last hundred feet or so. Grandma would have screwed that up, too.

Yea those 3 I am pretty sure would have been a fender bender at the least, a head-on at the worst.
Plenty of situations that I take over just because the truck is being inconsiderate, or a little goofy, those are expected.

 

[Zane Edwards]

FSD (supervised) is amazing technology.
But I feel like I have intervened and saved 3 accidents in the last 2,000 miles or so.
For the CT it seems like letting one driver for 100,000 miles without intervening would guarantee 1+ accidents, more less 1 million.
7million sounds extremely optimistic

hard to know, sometimes I intervene just because I am too close to something and don't want to risk it but other times when I am driving alone I let it do it's thing and haven't hit anything as yet. Unsupervised is going to be cool....really looking forward to my first RoboTaxi ride.

Sponsored