Consumers and AI

Consumer protection has been a key theme in public policy since the 1950s. Post-war affluence and technological innovation allowed consumers to access new products such as white goods for the kitchen and cars. These gave rise to safety and quality issues which were addressed in the US by a consumer movement headed by Ralph Nader.

Irish consumer law has developed under the benevolent influence of EU law that has introduced consumer protection under the guise of establishing the internal market. Consumer law has managed pretty well to address the issue of goods, but less so traditional services. Now, as we enter the Fourth Industrial Age with software and data becoming more important, consumer law faces new challenges to adapt so as to ensure adequate protection whilst leaving designers the ability to innovate. 

Products increasingly have software embedded in them (think of the GPS navigation system in your car) or connected to them in the Internet of Things as when heating systems or burglar alarm systems are controlled remotely by apps on your smart phone. The phone is itself an example of hardware housing numerous software applications. The EU has developed new rules to regulate the quality of digital content and services and has provided clear rules to determine when these apply or when the rules on sale of goods apply. The distinct rules on product liability (that govern harm caused by the product rather than defects in its inherent quality) are less clear. My view is that embedded software is already the responsibility of the final producer in product liability law, but clarity is needed about software liability to determine whether there is any independent liability for the software company. 

A few years ago I was involved in a project on lifetime contracts (see Life Time Contracts: Social Longterm Contracts in Labour, Tenancy and Consumer Credit Law, Nogler and Reifner eds). We focussed on employment, housing and credit contacts as long duration social contracts where there were changes in circumstances over time. However, many of these new consumer products embedding technology share these characteristics as they need to be updated to deal with bugs and changed circumstances. An interesting dimension is the reliance sometimes on data over which the service provider may or may not have control. Obligations to update have been introduced and provide a remarkable legal innovation by imposing a potentially long term post-contractual obligation. There has also been recognition of the need to take account of the ability to integrate into the operating environment. 

These new product developments can include the use of AI. Sometimes AI will simply involve software taking steps in line with pre-determined instructions. So a heating system app might be programmed to turn the system on when the air temperature drops to a set level or, more futuristically, when the app owner is within a certain time of reaching home. However, it gets really interesting where the AI has the ability to learn and make its own decisions based on data received from the environment. Automated cars have sensors that take in information about the environment and make decisions based on how they interpret the data. Some people might argue that the responsibility for this learned behaviour does not rest with the product designer whose liability stopped at the moment of supply, but it seems preferable to ensure they take responsibility for the risk they created by their initial programming of the AI entity. 

It seems the legal system can find answers to the liability rules. Some are already in place; others need a few tweaks and innovations to the established rules. However, consumers will still face an access-to-justice problem given the number of parties involved: who is responsible for the defect – the producer of the product, the software developer or the person who supplied data? In some instances, the solution might be to simply impose liability on one party as a convenient insurer. This is happening with automated cars. Producers are either accepting liability voluntarily to bolster confidence in the new technology, or no fault solutions are being found. The UK’s  Automated and Electric Vehicles Act 2018 makes the insurer liable, with the insurer then having recourse against the driver or car manufacturer. The assumption must be that car manufacturers will take responsibility where the car is at fault so as not to  be disadvantaged in the market if their cars become too expensive to insure. An alternative approach is to adopt network liability, which allows the consumer to sue any party in the chain, leaving the commercial parties to fight out where the cost should lie. For a more detailed discussion of this option see G.Howells and C.Twigg-Flesner, “Interconnectivity and Liability: AI and the Internet of Things” in Artificial Intelligence: Global Perspectives on Law & Ethics, Larry di Matteo et.al. forthcoming, available on SSRN). 

The EU has taken a healthy approach to the regulation of AI. It has promised rules on AI liability. Consumer protection in this area is valuable for not only consumers, but also for businesses if there is to be consumer confidence and acceptance of the exciting possibilities afforded by AI. Legal regulation is needed to support the acceptance of new technology. We need to make sure technology works in the interests of all of us as consumers; in similar vein see Frank Pasquale, The New Law of Robotics. I look forward to hearing Frank speak on campus in the near future.