Toyota announced that it plans to invest $1 billion in a Silicon Valley research center for artificial intelligence (November 6, 2015). On November 10, Volkswagen said it had hired away from Apple its lead expert on self-driving cars. (Yes, Apple too has a secret car project.) While analysts’ views differ on when, most agree that it is only a matter of time before fully autonomous vehicles become mainstream.

The US Department of Transportation called recent innovations by car manufacturers a “revolution in safety.” Historically, automakers (strongly encouraged by insurers) have focused on engineering vehicles to enhance occupant protection in the event of a crash. That’s why automobiles today have a range of airbags – front, rear, side and even curtains – as well as a long list of safety enhancements, including structural reinforcements to the passenger compartments and advanced safety belts.

Today, vehicle safety has expanded into technologies that help prevent or mitigate crashes. Vehicles can automatically brake to avoid or minimize accidents, self correct steering if the driver wanders out of his or her lane, and can parallel park better than many humans. They do this by means of a variety of sensors, connected to a central computer running sophisticated software. By use of sensors and cameras, today’s modern car can “see” round corners, keep a steady (and safe) distance from the vehicle in front, and anticipate and prevent a crash. All of these technologies, though, still require an attentive driver with hands on the wheel.

The technologies found in new cars today, together with the innovations car manufacturers are working on, will form the platforms on which self-driving cars operate. A report by the Boston Consulting Group found that almost half the world’s top 20 “most innovative companies” are automakers. For the first time, this year’s top 20 list included more car manufacturers than tech companies

By June of this year, Google’s self-driving cars had logged over 1 million miles on public roads, driving in fully autonomous mode. During the six years it took to achieve that mark, the cars were involved in 12 minor accidents, involving no injuries. According to Google, none of the accidents was the fault of the autonomous vehicle. Google appears to be in the lead, but traditional auto manufacturers are not standing still. BMW recently demonstrated a version of its electric i3 hunting through a parking garage to find an empty space, and then reversing into it. In January 2015, Mercedes’ luxury self-driving car took itself and some passengers to the Consumer Electronics Show in Las Vegas. In October, Daimler tested self-driving technology on a production line truck on a German autobahn. Chevrolet, Ford, Lexus, Audi, Volvo and Tesla are also apparently working on autonomous vehicles. Uber’s long-term plan is to operate a fleet of entirely robotic taxis.

As impressive as the technological advances are, challenges still exist. Google reports that its safety systems sometimes over-react when the car sees a potential danger. Additionally, self-driving cars always obey the rules of the road. That has lead to Google’s cars becoming stranded at a four-way stop intersection because the other drivers don’t actually come to a complete stop. Human drivers interact with each other visually, and a facial expression or nod of the head can inform others as to a driver’s intention. The engineers haven’t yet found a foolproof way to incorporate human behavior into a car’s programming. A recent article also wondered how a car might be programmed to make life-and-death choices, such as whether to hit a pedestrian or swerve and kill the car’s passenger.

Self-driving cars are coming because they offer the hope of dramatically reducing the enormous personal and social costs of accidents. The National Highway Traffic Safety Administration (“NHTSA”) reports that there were 5,687,000 motor vehicle crashes reported to police in the US in 2013. Of those, 2,310,000 were injury accidents and the rest were property damage only. 32,719 people were killed. The NHTSA estimates about 10 million additional crashes are unreported each year.

More than 90% of car accidents are caused by human error. Of the people killed, 10,076 died in alcohol-impaired driving crashes. 9,613 people were killed in speed related accidents. Other major causes of death and injury were red light running (around 900 people), fatigue (nearly 7,000), and distracted driving (nearly 3,000). The NHTSA estimates that in 2010 (the latest figures available), motor vehicle crashes cost almost $1 trillion in loss of life and loss of productivity – around $500 for every licensed driver in the US. Property damage claims in 2013 totaled around $18 billion.

All of these numbers have dropped noticeably since innovative crash avoidance systems – the precursors to fully automated cars – were introduced.

Insurance companies pay approximately 50% of all motor vehicle crash costs. Individuals pay 26% and health care providers and charities pay 14%. Federal revenues account for 6% and local municipalities and states pick up 3%. Overall, those not directly involved in crashes pay for nearly 75% of all crash costs, primarily through insurance premiums and taxes.

While one of the main goals of developing autonomous vehicles is to reduce crashes, accidents will still happen – they just won’t be the driver’s fault. Driver liability will likely give way to product liability as the leading cause of accidents. Sensors and communication devices will fail, software won’t respond, and the networks through which vehicles communicate with each other will go down. But what exposure will a company face – and what insurance and indemnity rights will it have – if one of its components fails and a massive multi-vehicle, multi-death accident occurs? What if a crash is caused not by a failed sensor, but because the sensor was broken by vandalism or an innocent non-driving related accident? Or had become covered in dirt or mud? What if a driver wasn’t trained properly to respond when the automated system failed?

As we transition into a world in which increasingly automated cars share the road with traditional cars, the mechanisms to sort out liability and pay for damages and injuries through insurance may get more complicated. In our next articles, we’ll look at how regulators envision their role as manufacturers seek to introduce more self-driving technology. Then we’ll outline the future insurance landscape for manufacturers and component suppliers, and for the continuing exposures of users.