An autonomous or self-driving car is one that does not require a driver to function. It can use motion sensors, cameras, audio receptors and advanced programming to move on its own, analyse environments around it, and take appropriate actions regarding the velocity and direction of its movement. In other words, it is a machine that do what most cars can do in the present day, but also perform the actions otherwise required of a driver. All one must do is direct the car to take them to a destination or travel along a certain route, and the car would drive itself, much like a computer code that prepares an output or result when given certain information.(1)  

Self-driving cars are part of a scale of six levels of automation, where each level has a varying degree of automated functions, with 0 possessing the lowest degree of automation and 5 possessing the highest. With the improvement in engineering and technology, each successive level of automation in cars becomes achievable. At present, semi-automated cars with features and requirements for manual human control are manufactured and most cars being sold fall under the bracket of levels 0-2 of automation. However, with how rapidly innovation is progressing, completely self-driven level 5 automated vehicles will be available within the decades to come.^[2] With their arrival imminent, the law needs to appropriately adapt to accommodate them without harming the interests of any involved stakeholders.

The significance of such vehicles comes from their revolutionary nature, which goes beyond anything humankind has seen in the field of locomotion.

Automated cars have two primary benefits among others. These cars are more inclusive in their access, as those who can't drive cars due to age or disability would have access to efficient means of commuting from place to place. Moreover, as they are automated and computer-operated, they are much less prone to making errors that human drivers do. This absence of human error would make them safer for use, as most motor accidents happen due to drivers' mistakes and lapses in concentration or judgement. Hence, self-driving cars would greatly reduce the possibility and occurrence of road accidents.^[3][4] Another benefit they provide is better traffic management.^[5] With speed management that adapts to the surroundings, efficient navigation technology, machine learning algorithms and vehicle-to-vehicle communication, self-driving cars can not only reduce travel times, but also reduce traffic congestion and optimise space utilisation on roads. ^[6]

With an impressive array of benefits, however, comes a list of concerns that cannot be ignored, and this is what makes them such an important topic of discourse. One of the primary concerns is that such cars would be extremely expensive and hence not be available to the general populace due to their lack of affordability. Moreover, with a complete and almost exclusive use of computer programming, various security issues would come into play, such as the susceptibility to hacking and security breaches. To add to this is the issue faced at the advent of the production of most machines- the loss of employment opportunities. For delivery personnel, post officers, taxi drivers and other people in the road transport industry, their jobs would effectively be replaced by machines that would be able to work more efficiently. From a more legal and ethical angle though, the issue of liability must be carefully considered. With automated cars, many legal questions arise with respect to the traditional road-related laws of negligence, accidents, and liability for mishaps.

Before self-driving cars are brought into widespread production and access, authorities around the world must ensure that the law is equipped to accommodate them. For this, one of the major legal principles that requires attention is that of negligence and the nature of liability.

Ordinarily, with cars that rely fully on driver inputs, the drivers hold responsibility and liability when their actions are the main causal factors for accidents or harm to other people or property. Examples include harm caused due to speeding, traffic light violations, driving while intoxicated, etc. In all such instances, the drivers' actions directly created a situation of danger for others and possibly themselves, and hence the drivers are said to be driving in a negligent manner, for which they are held liable.^[7] This is a straightforward way of imposing liability. Since the driver has control of the car, the driver should be held liable for errors leading to harm.

However, the issue of liability starts to become more complicated when the degree of control over the vehicular mechanics shifts from the driver to the car itself, such as is seen with self-driving cars. It would be natural to shift the liability on to the car itself with a higher degree of automation, as the car would be able to control itself. However, the imputation of liability upon the vehicle isn't simple. Even with automated cars, the driver can be held liable for negligence, if the car is at an intermediate level of automation where features like manual override are still present, and the driver doesn't take control even after being warned by the car to do so.^[8] With complete automation though, holding the drivers liable for negligence would be unfair to them as they don't control the actions of the car beyond simply setting a destination or route, and thus their own actions can't directly cause harm to others.^[9]

With respect to automated cars, 3 primary stakeholders come into play whom liability can be placed upon: the driver, the manufacturer of the car and the developer of the automation software.^[10] In order to then pinpoint the responsible party, the functions of each of the stakeholders must be defined and distinguished properly. For lower levels of automation, the driver's function would be to be aware of warnings and to control the car in situations requiring manual override. Failure in this function could lead to negligence. The developers of the car and the software could be held liable for negligence if they failed to take adequate care in manufacturing or maintaining their product, leading to inherent defects. Beyond that however, negligence doesn't come into play with self-driven cars, as the concept of reasonable care, being an inherently human attribute, would not apply to machines.^[11]

Just because self-driving cars cannot be negligent though, doesn't mean they can't cause harm or get involved in accidents. Despite their lack of human judgement and human error, they are still prone to 'machine error', or situations where the actions of the machine, although completely in line with the software, lead to undesirable outcomes. Illustrations of this include emergency scenarios where the car is forced to make a choice whether between protecting a passenger or a pedestrian, or between proceeding in one direction and harming some pedestrians or going in the other direction and harming others.^[12] Such situations don't have definite answers yet and researchers are looking for models which can clearly assign responsibility and pinpoint liability.

A popular model that is currently followed is mandating drivers to have control of the steering wheel. This assigns the driver greater responsibility and liability. But this model wouldn't be effective with fully automated cars as the driver would have almost no role to play. Other possible models include imposing strict liability on the owners of the cars, completely shifting the liability to the manufacturer, sharing liability between the two, and mandating insurance for road accidents.^[13] Some researchers have even suggested adopting the system used in the past for horse-based transport, as horses were intelligent beings prone to miscalculating and misunderstanding situations, leading to accidents, similar to automated cars.^[14] However. none of these models are perfect, and hence the quest for solutions continues.

There is a hope that the driverless technology will advance and develop to a stage where errors and accidents won't happen at all. But until that dream shows promise of becoming real, laws must adapt using various models and contingency plans. And if that is not possible, then innovation must wait for the law to catch up to it, for only then can innovation truly benefit humankind.

The author is a 2nd year student , West Bengal National University of Juridical Sciences (WBNUJS).Views are personal.

[1] Synopsys, What is an Autonomous Car, available at https://www.synopsys.com/automotive/what-is-autonomous-car.html (last visited on December 1, 2021).