What is the Relevant Product Market in AI?

Abstract

AI has taken the world by storm, and competition law is no exception. Policymakers, academics, and commentators are struggling to make sense of how to apply competition law principles to burgeoning AI markets. The question is spurred by an impending sense that inaction is likely to lead to monopolistic outcomes that will later be impossible to revert. What is feared is that AI will become dominated by a few large technology companies and, more spuriously, that these will be the same companies that already control vast swathes of the so-called digital sphere. In other words: it will make big tech even bigger.

One difficulty with this narrative, however, is that, strictly speaking, there is no such thing as an “AI market” because AI is not a unitary, monolithic technology. In this chapter, we argue that the first step to ensuring that antitrust law stays relevant in the age of AI is developing a principled approach to defining AI relevant markets; one that is legally, economically, and technologically sound. We suggest how this can be achieved by grasping the internal heterogeneity of AI and by understanding what makes AI similar to Human Intelligence powered tasks, thus eschewing simplistic narratives about AI’s supposed ubiquity and uniqueness that are bound to result in both the over and the under-estimation of relevant product market definition in antitrust law.