In September 2020, the Singapore Academy of Law (SAL) Law Reform Committee (LRC) published its third report under the series of “impact of robotics and artificial intelligence on the law”. This report, titled “The Attribution of Civil Liability for Accidents Involving Autonomous Cars”, considers challenges that automation in vehicles raises for laws, principles and practices applied for accidents and resultant civil liability.
Issues Arising in Accidents Involving Autonomous Cars and Current Situation in Singapore
Singapore has taken a light-touch approach to regulating autonomous vehicles, focusing on promoting innovation and testing of autonomous cars in “sandbox” situations rather than for mainstream use. Existing regulations touch on testing thresholds (regulatory authorisation and safety, maintenance and record keeping requirements) as well as liability coverage (mandatory insurance and failure alert system allowing the driver to take back manual control).
Once autonomous vehicles enter public mainstream use, the following 3 issues come to the forefront as the human driver element is gradually eliminated:
- For non-autonomous vehicles, the human driver responsible for the accident (due to his negligence) is required to compensate the other party. The compensation is covered by the negligent driver’s motor insurance.
- For self-driving cars, without a human driver, it will be difficult to determine how to split responsibility between the car manufacturer, manufacturer or a specific faulty component, and the owner or driver of the car.
- This depends on the liability framework chosen. There are difficulties for each framework, e.g. determining standard of care for negligence, or ascertaining fault and source of payout funding for product liability.
Possible Frameworks for Determining Liability
The report examines 3 liability frameworks:
- The main difficulty with using negligence to assess liability is determining whether any person/entity breached the standard of care, which would be pegged to industry standards or a reasonable manufacturer.
- Since autonomous vehicles usually will have multiple redundant and overlapping detection and safety functions acting together, the forensic process of determining the malfunction, as well as how much the driver was in control of the accident and taking proper care, would be complicated and costly.
- Product liability laws focus on whether there was a product defect, and whether the manufacturer’s failure to adopt reasonable product designs resulted in the harm.
- The primary difficulty with using even well-developed iterations of product liability laws (e.g. in the US) is the complex process of proving a defect existed with the vehicle’s software, and that it caused the accident. This is all the more difficult given complex less-than-transparent algorithms and varying training datasets used by the AI behind the autonomous vehicle software.
- A no-fault liability regime, where as long as harm was suffered due to an accident, the victim will get compensation from the manufacturer (without evidence of negligence or defect in product), may seem conceptually simple, but is rare and comes with its own questions.
- The question that follows is where the compensation is sourced from, and whether the no-fault liability will disincentivise manufacturers from entering the market or adopting high safety standards, since they will be automatically liable and draw from a compensation fund without having to assess standard of care or defects in products.
- In fact, in this case, the burden of screening out irresponsible manufacturers will fall on the government, which is another onerous burden for regulators to bear. This regime is also unsuitable for Singapore, as most manufacturers are based outside of Singapore and may resist any mandatory contributions to a common fund.
While the report concludes that it seems like keeping the existing negligence framework is most preferable, we can still discuss how to import desirable features of product liability and no-fault liability regimes for regulatory framework feasible for an “end state” of fully autonomous vehicles.
- The full report can be found here.
- Our Expert Guide on autonomous vehicles covering an overview of the liability landscape for autonomous vehicles for 21 jurisdictions can be found here.
- Our Law-Now article on the first two reports in this series can be found here.