By Ebere L. Ndubueze, Esq.[1]
INTRODUCTION
The proliferation of Artificial Intelligence (AI) in online business platforms is no longer news. Understandably, these platforms employ AI agents to simplify their processes while improving customer experience and cutting down on operational cost. The integration of AI into global marketplaces in the digital space has precipitated a paradigm shift in the manner of interaction between businesses and consumers. No longer confined to back-end logistics, AI now plays in the fundamental interface of consumption, driving everything from personalized product recommendations and dynamic pricing to creditworthiness assessments and automated customer support. For instance, it is not uncommon for a consumer to see advertised discounts for a gadget that they like pop up on their browser because they earlier did a Google search on the gadget or mentioned it on their social media.
However, with the benefits of this emerging technology comes a proportionate level of risks and actual harms to the customer who interacts with them. There is also the issue of this technology operating to dismantle traditional concepts of consumer autonomy, as seen in cases where a not-so-strong-willed consumer comes on a platform with an intention to buy some item and ends up buying another or other items after being swayed by their pop-up ads.
AI-DRIVEN HARMS In E-COMMERCE – case study of Moffatt v Air Canada
In the course of surfing e-commerce platforms, you might have relied on, or heard or even imagined that someone would rely on the advise given by a chatbot on the platform regarding the goods or/and services to be purchased. Now, take the imagination further to see that the customer believed and therefore acted on the instruction of the chatbot with respect to the services provided by the entity on whose platform the chatbot is. That is a typical example of reliance on AI-agents in e-commerce, and there can be no case of AI-driven harm without reliance on the AI-agent in the first place.
A decided case that clearly typifies this, is Moffatt v Air Canada[2] where the British Columbia Civil Resolution Tribunal found Air Canada liable in negligent misrepresentation for an incorrect advice given by the AI chatbot on its website. The facts of the case are that, Jake Moffatt who had recently lost his close family member wanted to buy flight tickets on Air Canada’s website for the bereavement trip to Toronto. He interacted with an AI chatbot on the airline’s online platform, which advised him of the available bereavement discount offered where the relevant form is completed within 90days after the ticket is issued, saying in essence that there is a discount whenever a flight ticket purchaser needs the ticket to travel because of death in the person’s family. The chatbot also directed Moffatt via hyperlink to a webpage of the Airline for more details on the bereavement policy. This webpage which Moffatt did not check, contained the accurate position of the bereavement policy to the effect that the discount did not apply to requests sent in after the trip had been completed.
What Moffat gathered from this AI chatbot’s representation, like any other reasonable person would, was that, the bereavement discount operated more like a refund if applied for within 90days of the ticket issuance. Believing this to be the representation of Air Canada, Moffatt acted on it and sent in his application for the bereavement discount within the 90days window. Unfortunately, Air Canada refused his application noting that the Chatbot’s representation was inaccurate. Aggrieved by Air Canada’s response, Moffat brought this lawsuit claiming negligent misrepresentation by Air Canada, in which the Tribunal ruled in his favour and awarded against Air Canada damages, pre-judgment interest and filing fees.
Interestingly, Air Canada argued that it should not be held liable for the acts or representations of its chatbot agent for reason that the chatbot was acting on its own accord, suggesting that the chatbot has a separate legal personality. Notably, the Tribunal rejected this line of reasoning as it amounted to stretching the principle of legal personality beyond reason. The Tribunal recognized that the airline owed Moffatt a duty of care which includes ensuring that its automated agents put out correct information about its business, as the training of these automated systems lied within the airline’s purview. This established duty of care was therefore breached when the chatbot misrepresented facts about Air Canada’s bereavement policy to Moffatt as a result of poor oversight functions by the Airline, and Moffatt suffered damage from relying on the misrepresentation, which reliance the Tribunal found reasonable. Hence, a case of negligence covering negligent misrepresentation was established.
Worthy of note is that, the tort of negligence is established where by the relationship existing between two people, one person owes the other person a duty of care and this duty of care is breached thereby leading to damages suffered, as established in the celebrated English case of Donoghue v. Stevenson[3] and restated in the Nigerian case of Universal Trust Bank (UTB) of Nigeria v. Fidelia Ozoemena.[4]
The reasoning of the Tribunal in Moffatt’s case is an affirmation that a corporate entity bears the liability for misrepresentations made by its agents or representatives including Artificial Intelligence agents or other automated systems on its online platform. This decision is in line with the legal principle of vicarious liability, whereby a Principal is considered liable for acts or ommisions of his Agent that occur in the course of the Agent’s duty as a representative of the Principal, as upheld in the case of Ifeanyi Chukwu (Osondu) Co. Ltd v. Soleh Boneh (Nig.) Ltd.[5]
THE NIGERIAN LEGAL RESPONSE
Where a case similar to Moffatt’s case comes up in Nigeria, the primary question will be on how the law determines liability. Besides the clear application of the tort of negligence such as negligent misrepresentation in such scenarios, the principle of vicarious liability will also set in whereby the court can interpret the Artificial Intelligence agent on the e-commerce platform to be an employee of the platform owner, considering that it is a system deployed and maintained by the platform owner and trained by it to serve its business purposes. Notably, vicarious liability can lie in cases where the employee’s or agent’s act or omission was negligent in nature provided it occurred in the course of discharging his authorized duties.
Further, the common law principle applying the control test can come to play in determining the extent of control exerted over the performance[6] of the AI system by the platform owner for the purpose of categorizing it as an employee, as opposed to an independent contractor. In light of this, where it is found that the platform owner controlled the output of the AI agent by overseeing its training and behavioral permutation, then it becomes clear that the AI system can be rightly categorized as an employee of the platform owner, not an independent contractor. For this reason, liability can lie on the platform owner by operation of the control test.
Statutorily, the CAMA[7] and FCCPA[8] provide interventions to protect consumers in this regard. By section 90 of the CAMA, a company is liable for acts of its agent where such was authorized by the members in general meeting (shareholders), board of directors or managing director. This underscores the principle of vicarious liability explained above because ultimately the shareholders and directors of a company are the mind and conscience of the company, therefore, a company’s decision to use an AI agent in its operations in many cases enjoy the sanction of the mind and conscience of the company. On its part, the FCCPA in section 123 prohibits the false or misleading representations to consumers by traders concerning their goods or services, while Section 125 provides that liability shall lie on any entity that engages in this false representation emphasizing that such entity shall be ordered to make monetary restitution.
CONCLUSION
The impact of AI on consumer protection is multifaceted, oscillating between the promise of empowerment and the peril of misinformation. Notwithstanding the advantages of Artificial Intelligence, large-scale integration of the technology comes with important legal and ethical challenges capable of undermining fundamental consumer rights. It remains a paradox that the technology which is greatly beneficial can be simultaneously and equally risky. Hence, it is imperative for businesses to take AI governance seriously by adopting the process of Human-In-The-Loop whereby there is not only human oversight over the decisions made by AI agents, but also active human involvement in the system’s workflow.
Considering the seeming equal measure of merits and demerits attributable to Artificial Intelligence, the need for regulatory measures to address the risks of this emerging technology cannot be overemphasized. Although experience has demonstrated that the law usually plays catch up with AI owing to the fast pace at which the technology develops, there is also need for a balance between encouraging the furtherance of innovation and consumer rights protection.
[1] Ebere L. Ndubueze is a commercial lawyer and LLM Cyberlaw candidate at the University of Lagos. She can be reached via email – hello.nellegal@gmail.com
[2] (2024) BCCRT 149.
[3] (1932) AC 562.
[4] (2007) 3 NWLR (Pt. 1022) 448.
[5] (2000) 5 NWLR (Pt. 656) 322.
[6] Okeowo v Migliore (1979) 11 SC 138.
[7] Companies and Allied Matters Act, 2020.
[8] Federal Competition and Consumer Protection Act, 2018.
The post Legal Liability For Ai-Driven Harms In Online Transactions appeared first on TheNigeriaLawyer.
