77 comments

[ 3.6 ms ] story [ 129 ms ] thread
Tesla assuming the financial liability for their cars may be what the self-driving industry needs for faster adoption.

One of the big questions with self driving cars is "who is at fault" and if Tesla the corp defacto volunteers by underwriting the liability, then that may pave the way for laws/policy aimed around the car manufacturers holding the blame.

That's not what this is. This is similar to what legacy auto insurers use to price risk using a mobile phone app or a telemetry device to gather data about your driving habits [1] [2] [3] [4] [5]. Tesla's primary use case was risk rating drivers for FSD beta access, but the secondary use case (or more likely, an adjacent use case with delayed deployment, never let a code path go to waste) is improving risk rating their insurance customers (good drivers pay less for insurance than bad drivers).

The differentiator here is that the functionality is built into the vehicle natively, and when on Autopilot, those miles are not included in your scoring (which aligns your interests with Tesla’s; they’re incentivized to make Autopilot safe and you’re incentivized to use it).

[1] https://www.statefarm.com/customer-care/download-mobile-apps...

[2] https://www.usaa.com/insurance/safedriving

[3] https://www.amfam.com/insurance/car/knowyourdrive

[4] https://www.allstate.com/drive-wise.aspx

[5] https://www.progressive.com/auto/discounts/snapshot/snapshot...

(disclosure: Tesla owner)

This is not what it is.

It's a regular insurance. Tesla isn't taking liability any more than Geico does.

When robotaxis happen, regardless of who makes them, the maker and/or operator will be liable for accidents.

There will be no big question.

Nothing else makes sense and probably nothing else could possibly be legal.

You can't say "you can go to sleep in a car but if it gets in an accident you'll pay for it".

That's why Tesla says you shouldn't use their autopilot while sleeping. They don't want be liable
I think you're assuming too much benevolence from Tesla here.

First of all, I think the way "FSD" is disclaimed (don't take your hands off the wheel, don't stop looking at the road) is going to mean that Tesla has a much easier time denying claims than a traditional insurance company.

If I'm, say, State Farm, and I'm insuring a Tesla that gets into an accident, I may not be able to prove that the driver was using "FSD" and/or whether they were holding the wheel, paying attention, etc.

Tesla, on the other hand, can get this data instantly and deny the claim based on very minor deviations from the "FSD" instructions.

Tesla controlling the data potentially proving their guilt (and liability as insurer of the guilty) is an obviously undesirable conflict of interests.

We already see this sort of conflict of interests in Kaiser Permanente health insurance and their hospitals, it's a huge problem.

Nice, welcome to the era of tailored fees based on assumptions rather than facts.

It used to be that you'd get a high insurance fee only if you cause incidents or damage.

But now not, you get to pay more if the Algo thinks that you drive shit.

Minority report is one step closer.

Since when? For a long time, it's been higher fees if you're in the wrong cohort, whether or not you in particular have been in an accident is secondary.
Not sure where you are from, but where I live is the way I described it.

There are factors such as driver experience, age and precious track record (good/bad driver based on accidents) but never based on driving style.

Right. How is your age or gender less of an "assumption" than your actual driving style?
Gender is not even in the picture.

Age is based on aggregated data which says that people with less experience are more likely to cause accident.

We can argue the whole day, but this is not a tailored fee as is not build based on my personal behavior.

Same reason that you can't drive a car when you are 5 years old or you can't drink before 18 (or 21).

> We can argue the whole day, but this is not a tailored fee as is not build based on my personal behavior.

You can argue all day, but it doesn't change the fact your main argument was you would no longer be charged based on facts but assumptions. The current method is largely based on assumption. Using actual data on your driving is not an assumption.

It may not have been your intention but that is how you wrote it. Two statements about no longer being charged based on facts, but on assumptions. One comment linking using your actual driving data as "assumption".

> welcome to the era of tailored fees based on assumptions rather than facts. > It used to be that you'd get a high insurance fee only if you cause incidents or damage.

> There are factors such as driver experience, age

So making assumptions based on "driver experience" and age is not making assumptions?

Is not a tailored assumption. If you can't see the problem I can keep pointing you to the moon the whole day, but you will still look at the finger.
Insurance companies have been offering discounts to install tracking and hard acceleration detectors for decades. None of this is new.

As a safe driver, I welcome these types of discounts based on good driving habits. If you do something associated with risk, like a sudden stop, then you should change your behavior.

Yes but was optional this is the difference.

Who does tell you that I drive dangerously in the road?

Maybe I bring my car in a circuit every now and then to have fun.

It’s optional to buy this insurance.
Yes, until is not. We're arguing about the direction not about the current state.

There will be a day when this would not be optional anymore

Why would no insurance keep offering this option?
(comment deleted)
To quote Warren Buffett: "If you've got a cop following you for 500 miles, you're going to get a ticket". When you have someone or something following you every second of every minute, with a perfect memory and no other context, your insurance is going up. And I'm sure some will say Tesla's AI can get context through other drivers actions; that's not quite right though - Tesla only knows what the other vehicles are doing, not the drivers.
Sure, but this isn't a cop. Insurance is a civil contract driven by market choice. This system isn't authoritarian - it's optinal and the customer has buying power.

Because of this, these systems typically are used exclusively to offer discounts against the baseline insurance premium and they can only lower the premium relative to the assumed default risk. It's all carrot, no stick.

https://www.nationwide.com/personal/insurance/auto/discounts... "Can SmartRide increase my premium? No. This is a discount program."

https://www.usaa.com/insurance/safedriving "Will my driving information impact my insurance rates? No. We will only use your driving information to calculate your discount."

Insurance companies are willing to offer these steep -- 40% -- discounts because bad driving is so extremely expensive and damaging. The added signal of driver behavior patterns can allow the insurance company to lower their rates by almost half. Almost half!

As an aside, these are the types of insurance premium reductions we should expect, or better, when self driving cars first become a reality.

Have you never bought insurance before? Car insurance at least has always been based on assumptions from a bunch of data about you.
Of course I did, but apparently seems like you're the only one having bought before right? I never paid more based on the driving style, simply because they no way to check.

There are overall considerations (eg. Age of the driver, cat accident history etc) but was never about how I drive my car.

You seem to be arguing against yourself?

You said "It used to be that you'd get a high insurance fee only if you cause incidents or damage".

I'm saying that no, Insurance cost has always been based on factors other than just accident history. And you seem to be agreeing with me.

Well, I was arguing about driving style. Don't read between the lines.

Driving style was never (at least from where I'm from) a factor that determines the fees of the insurance

If that's in the United States, then almost all states allow insurance to at least give you a rebate based on your driving style and the technology has existed for nearly a decade[0].

0: https://youtu.be/NalBLhboZQU

I don't live in the united states, but thanks for pointing that out.
(comment deleted)
So people should be priced based on the things outside their control rather than those they can control?
So what are you arguing for?

A) Insurance shouldn't know how often you hard-brake or tailgate (aka privacy)

B) Insurance shouldn't use the above data to "assume" how safe of a driver you are

Trying to imagine funny examples here: "Collision imminent! Your insurance is now $500 per second for the next 1.215 seconds"
Insurance already discriminates on age/car make based on shitty risk models, what's new
We could do better and not worst maybe
Well, this seems slightly better to me. I'd rather be discriminated against based on what I'm doing than on what I am. I'd rather not have my life dictated by unaccountable analysts/AI but I don't think this particular news is worse than the status quo.

EDIT: Main thing that irks me about this is the invasion of privacy. Less data = good in my book and this is the complete opposite so I'd probably pay a premium not to use that.

This is your personal choice, but you've to contextualize this in a larger framework which for most people is society.

So if there are lets say, 90% good driver and 10% of bad drivers, if you are happy to pay less (and you belong to the 90%) then you'd have a 10% people that either pay a lot for their insurance or don't pay it at all because it become too expensive.

So lets say out of the 10% drivers that have a bad behavior, 50% of them (thus 5% or the total) go out without insurance.

Now, it you are, for some reason, unlucky enough to meet one of the 5% without insurance and, unfortunately, you are on a car accident with one of them and that person run away without giving you a chance to ask for damage and, even if you can, he has nothing to repay off this damage, would you be happy?

Car insurance is required in Texas, so your hypothetical is somewhat nonsensical. If it's not required where you are from, maybe you should worry more about fixing that first?
It's required, but doesn't mean that there aren't people who do go on the roads without one.

Is hard to enforce the law to 100% of the population.

From where I'm from, the number of people driving without insurance has increased for various factors in the last few years. From a 0.x to a 1.x %

I think car make is reasonable thing at least for coverage of the car itself. More expensive and more expensive to repair models should have higher premiums. And they should not be subsidised by owners of cheaper models.
It is full self-driving, so the insurance should be free?
(comment deleted)
Can some explain the terms “safety score from trips” and “safety score for rating” as used in the example chart?
Everyone comparing this to minority report needs to calm down. Believe it or not, more accurate insurance premiums are a good thing.. if you are a good driver. And if there’s one thing that everyone agrees on, it’s that “I’m a good driver”.

If Tesla starts charging people who push the petal to the medal at every green light, or take every turn at 2G’s, then good drivers rejoice. I was a pricing actuary in a prior life, I would LOVE to see what kind of loss information they get from granular driving habits.

Accurate premiums are nice...if they don't require surveillance. Which this is.

Focusing on bean counting is missing the point.

In my mind the is absolutely the right kind of data collection we should be encouraging. Do I want Tesla to sell my driving behavior to 3rd party vendors? No. Do I want them to plug it into an algorithm and give me an insurance (hopefully) discount? Absolutely, 100x yes. Not ALL data collection is the evil eye of Sauron leaching off your personal life. Put it to good use, give us more accurate premiums!
> Do I want Tesla to sell my driving behavior to 3rd party vendors? No.

Am I confident that they won't? Also no.

This isn't a better use for data collection, we're just getting paid for the loss of privacy instead of giving it away for free.

> This isn't a better use for data collection, we're just getting paid for the loss of privacy instead of giving it away for free.

Name one single better use of my driving habit data than directly saving me money.

> one single better use of my driving habit data

Keeping it to myself.

Some people get paid to be erotic models; that's you. "Name one single better use of my hot bod than directly making me money."

Other people find it far more valuable to keep that to themselves.

And thus Tesla insurance is not the required insurance provider for Tesla cars, and enabling Tesla insurance telemetry requires clicking through a data-sharing prompt.
For now. And if "more data!" didn't seem to be the rallying cry of modern business, I wouldn't be so concerned.
Till all insurance will be like this and thus no choice for you anymore
So, it’s worth pointing out that Tesla already has a history of using telemetry against you, publicly, if you have the misfortune to die while using autopilot.
The question is how does it saves you money?

1. By increasing significantly the fee of whom is driving slightly worst than you (thus increasing the chance of having people without insurance driving cars)

2. By not allowing people at risk to purchase the insurance in the first place. Which increases again the chance of having people without insurance in the road.

This definitely helps insurance company to avoid paying compensations in case of accident so that they can insure only people that have very little chance to cause an accidental thus increase their margins

A huge step forward for insurance, a huge drawback for society.

Tesla has huge PR issues from media outlets jumping on things like autopilot failures as soon as the police report is publicly available. Selling insurance data wouldn't be something they'd benefit from if it meant more PR issues leading to fewer sales (as in, eventually, Tesla will plateau on sales to where every lost sale is important).
There was a time also when we thought Google would do no evil. When are we going to learn to just not trust big corps?
They are a good thing until they're not. Is already very bad for health insurance, with people being denied to have an insurance just because they're slightly overweight or they have pre existing conditions (like meh, is that the whole reason why I'm buying an insurance in the first place?)

Now fast forward 20 years and you can't buy a car insurance because you push too hard on the break.

> Is already very bad for health insurance, with people being denied to have an insurance just because they're slightly overweight or they have pre existing conditions

Are you talking about the US? Because the ACA made that illegal years ago.* Your slippery slope hand-wringing doesn’t hold water.

*Under the ACA, premiums can only be based on three factors, IIRC: age, sex, and whether you’re a smoker. (And zip code.)

Sex yes but weight no. Such a weird law.
Not in the US, but I've a close friend of mine who got rejected for an health insurance cause he's slightly overweight (and other people for preexisting conditions)

I'm not sure how that is in the US, but the abuse can come from anywhere and the result is likely to be worst that what they claim to be helping preventing.

> Health insurers can no longer charge more or deny coverage to you or your child because of a pre-existing health condition like asthma, diabetes, or cancer. They cannot limit benefits for that condition either. Once you have insurance, they can't refuse to cover treatment for your pre-existing condition.

I'm happy to see that is not the case in the US, but my was a general comment non necessary related to the US. But the fact that they're till law in the US presuppone that insurance company were behaving they way before.

What do you do if everybody rejects you for pre-existing conditions and you cannot buy an insurance? Do you live uninsured?
You can't (fully) control your health but you can control your driving style
...right until someone cuts in front of you or a kid runs after his ball onto street.
Those are exceptional cases. It's pretty easy to see when someone is frequently slamming on the brakes vs. occasional being forced to due to external factors.
(comment deleted)
I'm not a good driver. That's why I use Autopilot. Wonder what my premiums would look like. Technically I rarely drive these days.
> If Tesla starts charging people who push the petal to the medal at every green light, or take every turn at 2G’s, then good drivers rejoice.

The problem is that those things aren’t inherently what makes a good vs bad driver. If it was - are all race car drivers shit at driving? What about people who never do any of that and yet wreck their cars all the time?

Number one reason for people crashing is being distracted.

Insurance is only ever a good deal to those with bad habits or are unlucky. Truly Accurate insurance premiums become bad for everyone, as in a perfect information insurance system, you always pay in more than you take out.
Wouldn't not pressing pedal to metal at green light indicate a driver with very poor reaction times thus potentially higher risk in executing effective braking? And thus someone deserving higher premiums?
It's unrelated. A person might just want to get to the cruising speed so it doesn't need to constantly monitor the speed.
Yeah, not stopping at stop signs = higher insurance premium makes sense.
I'd hope that you could come up with something smarter based on some dynamic system that takes the data from people having accidents and scores you based on how close your driving behaviour is to the accident prone.
I used a similar service through Progressive nearly 2 decades ago. They sent you a dongle to plug into your OBD2 port and then measured how fast you accelerate / stop etc. It wasn't such a huge discount as to justify the surveillance I guess, but it was something.
How long until AI is capable enough to label a good from a bad driver ?
(comment deleted)
The comments in that article are very revealing. The “safety score” was probably developed by infallible ”data scientists” who know best and can whip up an algorithm in a weekend.