...not all cars with autonomous features necessarily operate on the same level of automation. |
Levels 0 through 2 require a human driver to monitor the driving environment, whereas that responsibility shifts to the automated driving system for levels 3 through 5.
Level 3 automation still requires human drivers, however, to respond to a request to intervene for "safety-critical functions".3
Level 4 does not require a human to intervene but a human is still necessary as the car's ability does not cover every driving scenario.
Level 5 is full automation where the car does absolutely everything and is expected to navigate every scenario in the same way a human would, including difficult conditions such as dirt roads or snowy conditions.
As a real world example, Tesla's autopilot system that is currently available in Canada is considered level 2 automation. It is capable of staying within the lanes, braking, accelerating and steering with the road, however, it requires a human driver to be attentive at all times and take over when appropriate.
While fully autonomous vehicles are not currently permitted to operate by law on Ontario's roads, (apart from pilot program testing), experts predict that we are only three or four years away from Level 5 production cars being widely available.4 That does not mean that they will be permitted on Ontario's roads, however, as regulations throughout the world (including Ontario) have consistently been outpaced by the technology. The main hurdle to the development of regulations is the question of liability in the event of an accident involving an autonomous car. While it is likely that liability will shift from the human driver to the autonomous system in fully autonomous level 5 cars, the pressing question right now is where the risk lies in the transition period from level 0 to level 5?
We know that under common law and current legislation, a driver can be held liable for damages for the negligent or reckless use or operation of a vehicle.5 In Ontario, an owner can also be held vicariously liable for the negligence of the driver and for failure to maintain the vehicle or for permitting someone to drive their vehicle who they knew was incompetent.6
The main hurdle to the development of regulations is the question of liability in the event of an accident involving an autonomous car. |
Applying those principles to the levels of automation, we do not see any reason why the same principles would not apply to all of them EXCEPT for level 5. Levels 0 to 4 all require some element of human input, however minimal, and accordingly create a potential basis for negligence against the driver. In level 4 for example, a human driver could be liable for negligently engaging the autonomous technology when it was not appropriate to do so given that the technology will not be equipped to handle all driving scenarios.
As for an owner's liability for failure to maintain their vehicle, we anticipate the current law will apply for all levels of automation, including level 5 as it will remain the owner's responsibility to ensure their vehicle is in proper working order, regardless of whether they are driving it. We anticipate owners' vicarious liability to be limited to levels 0 through to 4 given that it is only by reason of negligence in the operation of the vehicle that the statute is triggered, but that is something that will ultimately need to be clarified by legislation.
In fully autonomous mode, there is no human input and therefore no basis for a driver to attract liability in negligence. In Level 5, a driver is essentially rendered a passenger. It stands to reason that liability would shift to the manufacturer for producing a defective product. While we cannot say definitively that this is how the legislation will be framed, some of the leading manufacturers of autonomous cars themselves are assuming that it will be.
For instance, in October 2015, Volvo announced that it will accept all liability for accidents involving one of its cars while in fully autonomous mode.7 Google and Mercedes quickly followed suit with the same admission, all in an effort to settle the issue of liability and hasten the development of regulations. That essentially means those manufacturers will be held strictly liable by simple virtue of being involved in an accident regardless of who caused it. That of course also has significant benefits for insurers of the conventional cars that are involved in such accidents given that they will face much less exposure to tort claims even when liable.
Therefore, the question of liability when in level 5 autonomous mode appears to be relatively straightforward from our current vantage point. It is the process of getting to Level 5 from Level 0, 1 & 2 however, that is proving to be quite complex and there are a number of reasons why: 1) from a psychological perspective, as driver aids develop, drivers become less attentive. Features such as lane assist and blind spot assist have been invented and implemented in order to improve safety, and they do indeed accomplish that, but an unintended by-product of those features is that drivers rely on them and incrementally relinquish their involvement in the car's control. For example, Mercedes Benz has a feature called "Attention Assist", which is an option package that includes brake assist, lane assist and blind spot assist. The name itself implies that the vehicle will help the driver so they can pay less attention to the task of driving.
...the amount of time and attentiveness required by human drivers to take over in both critical and non-critical situations, creates potentially significant risks |
This conflict between human attentiveness and developing safety features has even attracted tension between some of the major manufacturers. For instance, Elon Musk was criticized by Google, Ford and Volvo for pushing the development of Level 3 autonomy, citing concerns about the transfer of controls between the autonomous system and the human driver.8 The concern is that the amount of time and attentiveness required by human drivers to take over in both critical and non-critical situations, creates potentially significant risks.
Another layer of complexity during the transition phase is the question of liability surrounding the various third party software and components involved in autonomous cars. The failure of a component within the car, such as a sensor or camera, will likely be caught up in the current regime of product liability law that will apportion an appropriate percentage of liability to the car manufacturer. But what about technology outside of the car? The paradox about autonomous cars is that they actually become less autonomous as they become more autonomous because they become increasingly dependent on third party technology. For instance, apart from communicating with satellites for global positioning, autonomous cars will increasingly rely on two other significant forms of communication: V2V and V2I.
V2V stands for "vehicle to vehicle" communication. It is a network of communication between automobiles wherein they send messages to each other with information about what they are doing. V2I stands for "vehicle to infrastructure" communication. It is a network of communication between vehicles and infrastructure, such as traffic lights and signs whereby they communicate with each other in real time9 and it is part of an emerging technological trend called the "Internet of Things" that involves the inter-connectivity of devices through the internet.10
Vehicle to vehicle communication is expected to be more effective than current systems of lane departure, blind spot detection and cameras because it creates a real time 360 degree awareness of surrounding risks.11 Vehicle to infrastructure communication is intended to not only improve traffic flow but also include key safety features such as alerting drivers to hazardous situations up ahead including traffic congestion and local weather systems. V2I technology is projected to be ubiquitous within a few years as cities across the world work toward becoming "Smart Cities" by increasing their inter-connectivity.
We anticipate that as the autonomous technology develops along with V2V and V2I technology, litigants will be looking to new sources of liability, such as third parties that design and maintain V2I technology, companies that provide cars with live global positioning data, weather data, or professionals that maintain and service the software of autonomous cars. The success of such claims will largely depend on how the courts decide to deal with the issue of defective information, specifically defective information that is provided to autonomous cars that is discovered to be the cause or contributor of a loss. Is information considered a product that can fall under product liability law or will an entirely new category need to be created? While we cannot answer that question quite yet, we can say that, absent legislation, courts will be left to apportion liability among the various actors who have caused or contributed to the loss or damage just as they do under the current tort regime. Overall for litigation, this will mean more parties involved in actions and more sophisticated methods of obtaining evidence and determining liability will be required.
Finally, it is no secret that the future of the automotive industry and the insurance industry will largely be driven by data. The interconnectivity of cars and infrastructure will rely on an expanding collection of information that will become increasingly useful and valuable. That of course raises the major issue of cybersecurity. This interconnectivity and proliferation of data will make our personal data vulnerable to hacking and will also make vehicles themselves vulnerable.
It would appear that with each technological innovation and advancement comes a new cyber security risk |
The synchronization of phones to car systems means hackers could access a driver's banking information or personal correspondence by gaining access to the car's network. Hackers have also proven an ability to physically lock and unlock cars remotely, cut the power to engines, disable brakes or engage the throttle.12 It would appear that with each technological innovation and advancement comes a new cyber security risk. As far as the vulnerabilities of the cars themselves, we expect liability to fall on the manufacturer, whereas for vulnerabilities related to third party sources (such as V2I infrastructure) we anticipate liability will fall on the municipality and/or maintenance provider.
Therefore, while the transition phase towards level 5 automation threatens to be complex and costly, we are confident that Ontario's tort system is capable of responding to it. Level 5 autonomy will remain elusive for the time being, subject to the development of regulation and legislation, but it remains poised to offer a simpler system overall for dealing with liability in auto claims.
1 A Pilot Project to Safely Test Autonomous Vehicles – Summary of Proposal
2 SAE International, "Automated Driving, SAE International Standard J3016"
3 Autonomous driving levels 0 to 5: Understanding the differences Hope Reese, January 20, 2016 Tech Republic.
4 Olivier Garett, 10 Million Self-Driving Cars Will Hit The Road by 2020-Here's How to Profit Business Insider (March 3, 2017)
5 Highway Traffic Act, R.S.O. 1990, c. H.8, s. 192(1).
6 Highway Traffic Act, R.S.O. 1990, c. H.8, s. 192(2).
7 Kirsten Korosec, Volvo CEO: We will accept all liability when our cars are in autonomous mode Fortune (October 7, 2015)
8Fred Lambert, Elon Musk defends level 3 autonomy against Google & Volvo Electrek (September 13, 2016). .
9 Margaret Rouse, Vehicle to Vehicle Communication Internet of Things Agenda, (October 2014).
10Andrew Meola, What is the Internet of Things Business Insider (December 19, 2016).
11 Supra note 9. .
12 Viereckl, Ahlemann, Koster and Jursch, Connected Car Study 2015: "Racing Ahead With Autonomous Cars and Digital Innovation", p.30.
