Vedant Saxena is a LLM graduated student of National Law University, Delhi
Introduction
The dawn of AI-assisted online retail has introduced a fresh spectacle to the debate on whether Artificial Intelligence (AI) will be able to replace the human mind. While AI cannot take over the purchasing power of the consumer unless such power is specifically delegated to it by the consumer, it can still act as a personal shopper so long as it possesses information about the products on sale. On account of this, the very basis of trademark law, i.e., the identification of a specific good or service, seems to have been rendered obsolete. For instance, in ‘target advertising’, personalised recommendations are provided to customers, based on their web searches and prior customer purchases and other external factors like speed of delivery, availability and nutritional information. Such recommendations are, therefore, mostly irrespective of the brand. With the growing dominance of AI, this could mean a gradual erosion of the role of branding and, thereby, trademark law. Through this article, the author assesses whether personalised AI shoppers threaten to dilute the very foundation of trademark law by introducing data-based predictions against brand-based consumer choice.
Tracing the cyclical evolution of the purchasing process structure
For understanding the impact of AI on trademark law, it is essential to trace the evolution of the purchasing process itself. The traditional structure of buying and selling products was centred around ‘shopping assistants’, who provided a ‘filter’ between the product and the consumer. These shopping assistants possessed extensive knowledge of the products on display in order to aid consumer choice. With both limited products and personalised assistance, consumer confusion, which eventually led to the birth of trademark law, was a relatively minor concern. Trademark law witnessed a significant boom with the advent of the modern supermarket, where consumer choice was rendered independent of any external assistance. In light of this, the possibility of consumer confusion was increased manifold. In the absence of a filter, it was now the brand speaking to the consumer. With businesses now relying on consumer recognition and brand identity to regulate sales, trademark law gained prominence.
However, today, with digital transformation at the forefront, an intermediary has once again been introduced, albeit an artificial one. The growing popularity of e-commerce platforms, personalised AI recommendations and virtual assistants has marked the return of the shopping assistant. However, such shopping assistants exist as intelligent algorithms, which are fed extensive volumes of consumer data, propose options and behavioural patterns to predict consumer choice. With the operation of such assistants, consumers appear to be following algorithmic patterns rather than making independent decisions based on brand value.
Decoding the shift from ‘Branding’ to ‘Personalised Product Recommendations’
As has been explained before, with the advent of the modern supermarket, consumer choice was rendered independent of any external assistance. In light of this, consumers relied on brand value to adjudge their preferences. An increased reliance on brand value, therefore, allowed businesses to cultivate goodwill. However, with branding came deception, i.e., the will to wrongfully use another entity’s trademark to encash upon the latter’s goodwill. In order to prevent maliciousness, trademark law was formulated. A trademark usually refers to a word or device, which is employed by a user to indicate to the consumers that a particular good or service originates from him. Using deceptively similar marks or such marks that do not serve to distinguish the consumer’s goods or services with another’s, is prohibited.
With the advent of AI-assisted online retail, the premise that consumers rely on brand value to purchase products is challenged. Several AI applications, such as Amazon Alexa, AI robot assistants, such as Pepper, and AI personal shopping assistants, such as Amazon Dash and Mona have been introduced, which have significantly altered the purchasing process structure. These shopping assistants rely on demographic information, purchase records and search histories to provide personalised recommendations to consumers. Therefore, while trademark law aims at utilising human memory to help businesses thrive, AI-assisted online retail causes the consumer to be spoon-fed, thereby diluting the very basis of trademark law. A significant example of this paradigm shift is target advertising that personalises consumer experiences through accurate recommendations. This involves consumers receiving product suggestions based on their previous activity. Moreover, there is empirical evidence indicating that e-commerce websites providing personalised recommendations are likely to be preferred over websites which do not provide such. Through a study published on Future Business Journal, it was found that upon being used as a moderator, the presence of personalised recommendations made the model 5% more accurate in explaining customer satisfaction. Several product recommendation engines also provide suggestions such as ‘Customers also buy’ and ‘Often bought together’ to first-time consumers, based on the data of users who have made purchases from the website. Therefore, the aim is not to generate personalised recommendations based on the goodwill of the brand, but to do so based on what the consumer would most likely buy next. With the consumer’s choice being regulated by AI, the purchasing process structure appears to be returning to the pre-supermarket era, where branding and consumer confusion were minor concerns.
In his book titled “Prediction Machines: The Simple Economics of Artificial Intelligence,” Mr Ajay Agarwal explains how the prevalence of such practice would cause a substantial deviation in the present approach. With the current mode of online shopping following the pattern of ‘Shopping to Shipping,’ the approach would soon be ‘Shipping to Shopping’. In simple terms, in the present day, people usually select amongst their favourite brands and the product is subsequently shipped, but with the advent of AI in online retailing, the products shortlisted via AI, based on buying profiles, preferences or search histories, are directly shipped to the consumer who subsequently finalises the payment based upon his liking.
Deciphering the implications for trademark law
With branding at the forefront before the dawn of predictive AI, the requirement of a legislation preventing unauthorised uses of registered or well-known trademarks was indispensable. Trademark law, as has been explained before, aims at ensuring both that owners are adequately rewarded and consumers are not deceived by deceptively similar trademarks. However, with online retail now increasingly being predictive rather than brand-driven, key doctrines within trademark law need to be re-examined. Certain terminologies, such as ‘likelihood of confusion’, ‘average consumer’, ‘secondary infringement’ and ‘imperfect recollection’, which are essential to assessing infringement, are becoming redundant with the prevalence of AI. Such terminologies were formulated in the context of human beings. For instance, the question of trademark infringement is usually decided based on the judgement of an average consumer, who is presumed to possess ordinary intelligence and imperfect recollection. If it is established that such a person would confuse the alleged infringing mark with the original, infringement would be deemed to have been proven. However, in the event of AI intermediating the purchasing process, whether in the form of generating personalised recommendations or filtering results, the concerned decision-maker is no longer human. Therefore, with predictive AI leading the purchasing process structure, consumer confusion appears to have been rendered obsolete.
However, on the contrary, the rise of AI may enhance the effectiveness of trademark enforcement. Charlie Hill, the head of the product division of a global IP resource organisation, namely ‘Trademark Now’, stated that AI is more skilled in drawing comparisons between trademarks. There are several factors, such as imperfect analysis by the human eye, imperfect recollection, unsynchronised trademark offices and numerous trademark registrations, which render a human analysis flawed. With the prevalence of AI, therefore, such issues may be resolved. However, with the prevalence of personalised AI assistants, it is imperative that trademark law is revamped to accommodate algorithmic influence. In light of the fact that AI-driven systems distort market choice, courts and legislatures would need to develop novel standards for accountability and review thresholds for confusion.
Conclusion
With AI taking the world by storm, it is an undisputed fact that almost every facet of the legal field will be affected, with trademark law being no exception. With the return of personalised shopping assistants, albeit in the form of artificial entities working upon algorithmic patterns, legal doctrines built upon human cognition, such as ‘likelihood of confusion’ and ‘imperfect recollection’, are increasingly losing relevance. However, as has been explained above, with AI being more skilled in drawing comparisons, it may be utilised to enhance trademark enforcement. The prevalence of AI applications in the field of online retailing, therefore, calls for a more advanced approach to handling cases on trademark prosecution and litigation. It is therefore imperative that Indian jurisprudence considers such dynamic interpretations to ensure the peaceful co-existence of AI and trademark law.