Publication Date

2026

Abstract

This chapter identifies three common design choices that must be made in the course of an effort to regulate artificial intelligence “AI”). By isolating the way in which those different sovereign regulators approach these choices in respect to AI, it is possible to evaluate more clearly the likely effects of regulation, and to better understand aspects of regulatory design that to date have been overlooked. First, the “what” of AI regulation is not just a matter of how “AI” is defined, but also which element of the technology stack is targeted by law: compute, data, or models. Second, the “who” of regulation concerns the identity of the regulator. For a sovereign- state regulator can often delegate regulatory tasks down to subnational units, or else up to a transnational body. Third, AI regulation itself can have a number of different formal characteristics as law. While there is an extensive literature on regulatory form, certain key choices in the AI context have not received as much attention. By asking “what,” “who”, and “how” of AI regulation, it is possible to grasp the central choices that any given sovereign makes, and hence map the likely effects of or limits of a regulatory project. The chapter directs particular attention to the questions of “what” and “who” because these are less studied, and hence less well understood than the question of “how.” Put together, though, the chapter’s analysis of all three design margins offers a comprehensive accounting of the overall landscape of, and challenges to, AI regulation.

Number

26-4

Download


Included in

Law Commons

Share

COinS