As self-driving cars become more popular in mainstream society,they're also raising lots of questions about safety. One of themore interesting involves the inevitable accidents and how blame will be assigned. This is importantbecause in the car insurance industry, blame helps determine whichinsurance company pays for damages.

Related: Uber autonomous car involved in fatal crash inArizona

Is a human driver free of blame if a vehicle was on autopilot atthe time of a crash? Does the maker of the car assume liability? Orshould the driver or passenger be expected to step in and resolveany issues before they result in an accident? Is there anyprecedent for this sort of thing? Let's take a look.

Automakers are confident

Automakers and the tech companies behind self-driving cars areextremely confident in their creations. Tesla, one of the mostfamous proponents of autonomous vehicles, is so sure of the safetyof its vehicles that it has started offering insurance for life for purchasers ofnew Tesla vehicles. Waymo, a division of Google's parent companyAlphabet, also recently announced a similar initiative,providing insurance free of cost to passengers in Waymo'sautonomous ride-hailing vehicles.

Needless to say, for any business to offer up an incentive likethat, it has to be confident that the risk is minimal. And, so far,that's proven to be the case. There have been a handful of mishaps— including a fatal accident involving a Tesla a few yearsago — but for the most part, self-driving cars have proven to beexceptionally safe.

What makes autonomous cars so safe?

Traffic accidents, unfortunately, are a common cause of injury,and they're expected to become the fifth leading cause of death by 2030 unlessaction is taken. Experts generally agree that a large majority ofaccidents are caused by preventable human error.

Someone takes their eyes off the road for a moment, forgets tolook before accelerating, doesn't notice a red light, getsdistracted by the kids in the backseat — these things happen, andif you're in the wrong place at the wrong time, they can easilylead to an accident.

AI is far less vulnerable to these types of problems. Aself-driving vehicle always checks all directions for oncomingtraffic, always has its “eyes” on the road, and doesn't getdistracted (the big one). With these issues solved, theself-driving car can safely head on its way, while passengers arefree to relax or get started on their work.

Expect some bumps in the road

Not all is perfect in AI-land, though. There are situations AIcan't easily “think” through, presenting head-scratching challengesto engineers. These tricky situations tend to have one thing incommon: they present unexpected circumstances to the self-drivingAI.

First and maybe most dangerous is reckless drivers. Unexpectedbehavior from other (human) drivers on the road may throw AI for aloop, and it's difficult to program for every possible poor choiceother drivers might make. For example, if someone passes in ano-passing zone and narrowly misses oncoming traffic, how shouldthe self-driving vehicle respond?

Humans recognize these situations and can adjust accordingly,and it's likely that AI will eventually recognize them as well, butit may take time, since the self-driving vehicles need to beprogrammed to recognize every possible permutation of thesescenarios.

There's also uncertainty around human oversight:

• Should self-driving cars be required to have passengers capableof taking over in dangerous situations, or should we trust that acomputer can react faster and make a better call than a humandriver?
• Could a human driver trying to take over actually slow down allpossible reactions and make the situation worse?
• And if people are expected to cover for AI in trickysituations, should they be held liable if they don't?

Fault may be tricky to place — as it always hasbeen

All these unknowns are why it's also up in the air as to who —or what — would get blame in these situations. But this isn'texactly a new phenomenon. Determining fault in an accident hasalways been complicated, and autonomous vehicles won't changethat.

“With regards to the fault of an accident, even today insurancecompanies fight tooth and nail to discover the fault of theaccident,” says Kevin Courtright, licensed insurance expert withTheGeneral Auto Insurance. “Different states have different lawsto not only assign fault in the accident but also to determine whatpercentage of fault each party should assume. This is done in aneffort to fairly treat both parties. If it is determined that theaccident was only 65% [the fault] of driver A, why should he or sheneed to cover 100% of the damage to driver B and his car?”

It's possible that insurers could work out a similar,percent-based system to assign fault when AI enters the mix, thoughfuture legislation could force another solution altogether. Maybeall self-driving cars will end up with little airplane-type blackboxes to record everything leading up to an accident.

There's an exciting future ahead

Ultimately, there are a lot of questions stillunanswered about self-driving cars, safety practices, and insuranceclaims. And the transition to automated transportation willlikely be a little touch-and-go for this very reason. Many of thesequestions will have to be figured out along the way, and many moreare sure to pop up in the process.

We want to live like the Jetsons — reading a book, watching a TVshow, or taking a nap on our way to work. But we haven't evendiscovered all the baggage that autonomous vehicles will bring withthem.

There's little doubt that as the technology is perfected, manyof the safety concerns surrounding autonomous cars will be quelled.The next big problem will be to determine how damages will becovered. So if you're in the insurance industry, this is definitelythe space to watch right now.

How these scenarios play out on the road will have far-reachingconsequences for how insurance claims are handled in thefuture.

Parker O'Very ([email protected]) is adigital journalist who writes about automation, insurancetechnology and consumer tech.