When brand distinctiveness is in the AI of the beholder: trademark law in the age of AI shoppers

Trademark law is quintessentially a response to all too human cognitive failures: We have limited time, limited attention spans, limited memory and an ability for pattern recognition that on the one hand is highly evolved, but also prone to manipulation. AIs are (potentially) not encumbered by these. They are faster, do not tire, have potentially unlimited access to data and while their ability to recognize patterns may not yet be on a par with humans, they are also less likely to fall at least for some forms of deception. As we increasingly rely on AI to make purchasing decisions (recommender systems), communicate our purchase decision to an AI as intermediary Alexa) or even leave it entirely to autonomous agents, a number of difficult questions arise for our trademark regime. The first set of questions centres around the notion of confusion. If a product’s use of an identifying sign would not confuse a human, but might well confuse an AI shopper, is this infringing behaviour? Is the “average consumer” now a consumer that also uses AIs, or should we go even further an include AIs among the group of “average consumers”? Attacks against neural networks can lead to intentional misidentifications of patterns. In addition to possible criminal charges, are there trademark issues in such an attack? Since robots can sense in ways very different to humans, is there an argument to be made that “non-traditional trademarks” will be of increasing importance and in need of more systematic international protection? We will look in particular at the interaction between non-traditional trademarks and attacks against machine learning to elucidate some of the more complex issues that the new technology brings to the regulatory landscape. As AI supported shopping agents can work across national borders, we ask if this too points towards greater need for international harmonization.