Tesla Autopilot is 8 Times Safer Than Regular US Cars

Tesla electric cars are over 8 times safer than regular US cars. Tesla’s vehicles have had the lowest probability of injury of any vehicle tested by NHTSA since 2011. Recent safety data shows Tesla’s on Autopilot got into 1 crash every 4.34 million miles, versus the average US driver who got into a crash every 498K miles. Tesla electric cars are 8-9X safer than the national average. Tesla’s self-driving tech is almost an order of magnitude safer than human drivers and constantly improving.

17 thoughts on “Tesla Autopilot is 8 Times Safer Than Regular US Cars”

  1. If you look at the graph it shows teslas without auto-pilot as well. They finish up at about 1.8 million miles for each accident, compared to auto-pilots 4.3 million miles.

    Although there is still a question mark about if the driving conditions are a fair comparison.

  2. I interpreted that graph as a combination of four things.
    1) A slight improvement year on year for the Tesla auto systems.
    2) A big dip in the winter months for the Teslas. This is North America which is close to the Arctic and has dreadful weather during the 4th and 1st quarters which are their winter. Clearly this bad weather doesn’t suit the Teslas.
    3) As everyone (except the PR person at Tesla) points out, the autopilot is only used on open highways in good conditions where everything is safer per km anyway.
    4) As Jayco points out, these are expensive, new cars. The driver population will probably be a lot safer than those driving $500 rust buckets from 1973.

  3. That’s a great point. Look at IIHS death numbers by million registered miles and note car models that are only very slightly different (for example wagon vs. sedan) where the sedan version is marketed to young drivers. The sedans have a massively different (higher) death rate.

    The difference? The driver and passenger demographic.

  4. There’s a lot less Tesla vehicles in Q3 2018 is my guess; way more statistical noise. I’d ignore that quarter and the one after.

  5. You’re 2.5x less likely to crash/die per mile on the highway. Autopilot is just about 100% used on highways.

    So 50/50 highway non-highway is 1.75x modifier blended.

    So don’t forget to divide the Autopilot numbers by that amount.

  6. I’ve heard the medical profession is concerned about the impact on organ transplant availability in the fairly near future, as the large majority of the “material” comes from traffic accidents.

    Getting transplants already involves substantial waiting lists. Now reduce availability by “an order of magnitude.”

    Not to say that we shouldn’t eliminate every traffic accident we can! We absolutely should. But it’s all the more reason to put increased focus on growing (or printing) new organs, and regenerating existing ones, where possible.

  7. What happened before Q3 2018 to deflect the numbers down or in Q1 and Q2 2019 to produce an upward deflection? Is there a story there that says something about these numbers? Model 3s are entering the mix in this period. Are Model 3 owners more accident prone than S&X owners as a subgroup? It’s a possibility. Even so, it seems likely that Tesla introduced features are bending the data up after this.

    It would be interesting to see much more detailed data or even have raw data made public (without any personal identifiers). Tesla has better data about safety performance than anybody else in history and it will influence public opinion and regulators on autonomy.

  8. Given the auto pilot can only drive in the easiest conditions it’s not that great of a comparison. Maybe drivers have even less accidents in those conditions, the data would have to be broken up to tell.

  9. I agree with the earlier commentators that the numbers need more context in order to provide a meaningful comparison. One of which I would like to see is comparing the lower level driving assistance feature performance with their competitors systems – like what’s in the Subaru I bought my wife earlier this year.

    Without the context, these numbers are nothing more than a sales pitch

  10. Not an apple-to-apple comparison, have they restricted the data of US average to the same kinds of routes and conditions a Tesla auto-pilot takes ?

  11. I do love my m3. But it is probably safe to say 95% of Tesla’s are owned by people 30-65. Not 16 year olds who can barely drive nor 95 year olds that probably shouldn’t be. They also are probably not alcoholics who constantly drink and drive.

    It sounds great but probably not comparing apples to apples.

  12. It’s all bullshit because in dangerous conditions ( like heavy rain or fog) you need to turn off autopilot and drive manually. So accidents usually happen during manual driving.

  13. My thoughts exactly. As I understand it, Tesla autopilot is mainly used on the freeway, so if there were numbers of freeway accidents of “manual” drivers, then that would be the relevant comparison.

    Fortunately, we may soon have an apples-to-apples comparison. Elon Musk has claimed that full self driving (feature complete) will be available to a limited number of users at the end of 2019. Lets keep our fingers crossed that he can deliver on his word!

Comments are closed.