Regulating competition in generative AI: A matter of trajectory, timing and tools

Friso Bostoen, Anouk van der Veer

Research output: Contribution to journalArticleScientific

Abstract

In January 2024, the European Parliament called on the European Commission (EC) to consider adding generative AI (GenAI) services to the list of “core platform services” of the Digital Markets Act (DMA). Others argue that the European Union (EU) need not wait for new regulation: competition law’s abuse of dominance provision can already be used to tackle concentration in the industry. To help shape its views, the EC launched a call for contributions on competition (or the lack thereof) in GenAI markets.

This is serious regulatory zeal for a technology that only became widely used at the end of 2022. But the present must be understood in light of the past. The previous platform shift, in the 2000s, saw the rise of dominant players in key Web 2.0 markets like search and app distribution. The prevailing idea is that regulatory intervention came too late to keep those markets effectively competitive. Hence, it is better to intervene early now.

But is the analogy justified? Is GenAI a platform technology that requires immediate regulatory scrutiny to stay competitive? To find out, we address three sub- questions. The first concerns the trajectory of GenAI: is the market destined to evolve into a few ecosystems ruled by dominant platforms? (I.) The second concerns the timing of regulation: given the new and quickly evolving nature of GenAI, what is the “right” time to act? (II.) The third concerns the regulatory tools: can we put existing competition law to work, or do the market dynamics call for a tailored instrument? (III.)
Original languageEnglish
Article number118602
Pages (from-to)27-33
Number of pages7
JournalConcurrences
Volume2024
Issue number2
Publication statusPublished - May 2024

Fingerprint

Dive into the research topics of 'Regulating competition in generative AI: A matter of trajectory, timing and tools'. Together they form a unique fingerprint.

Cite this