Not “big is bad” but “closed is bad”? Exploring dynamic competition in generative AI

Scholars have seized upon studies on economic concentration to revive the idea that ‘big is bad’. A recent addition to the literature, by James Bessen, shares a concern for declining disruption rates but argues instead that ‘closed is bad’. We review this debate, concluding that neither fully grapples with the dynamic competition paradigm. The theory of dynamic competition offers important nuances to current understandings of competition, and in particular the role of innovation, which we illustrate via a case study of generative AI. We then turn to policy measures, and in particular those under the ‘unbundling’ umbrella, i.e., breaking ties, mandating interoperability, and imposing data sharing. Each of these measures is finding their way into the EU policy framework. We show that, though such measures can make sense in specific circumstances, the challenges—both conceptual and practical—are not to be underestimated. We build on our generative AI case study to highlight the opportunities and risks involved.