An American's Guide to the EU AI Act

Abstract

The EU AI Act entered into force in August 2024. The AI Act is long. It is complicated. It relies on a regulatory framework and institutions unfamiliar to many in the United States. But as the first omnibus AI regulation worldwide, it has the potential to have a vast influence on both practice and lawmaking.

In this Article, we provide the American’s Guide to the EU AI Act. This Article breaks down the AI Act for a U.S. law audience, explaining the overall mechanisms, and how the Act interacts with background EU laws and institutions. At its core, the AI Act is structured on Europe’s product safety regime. It is aimed at governing AI systems through assigning them into risk tiers, and deploying bans, risk regulation, and self-regulation. But it also contains later-drafted provisions on general-purpose AI that depart from this framework, as well as multiple ad-hoc provisions and other regulatory strands.

The Article also analyzes the consequences of framing AI regulation as risk regulation and of constructing AI systems as products rather than bureaucratic processes. It describes the AI Act itself as a “legal exoskeleton,” with hard law built around the softer belly of technical standards. The Article identifies the threats this poses for both substance and legitimacy, and the potential political ramifications of that design.