The EU AI Act extends its jurisdiction beyond EU borders, applying to any company whose AI systems or outputs are used within the single market. Five papers examine whether this scope produces a Brussels Effect, the phenomenon, first theorised by Anu Bradford (2012), whereby EU regulatory standards are voluntarily adopted by foreign companies and governments due to the size and attractiveness of the EU market.
Siegmann & Anderljung (2022) distinguish a de facto effect, whereby companies extend EU-compliant products globally rather than maintaining separate compliance regimes, from a de jure effect, whereby third-country governments adopt analogous legislation. They conclude both are likely, particularly for large US technology companies operating high-risk AI systems as defined by the Act.
The Brookings Institution (2023) with Alex Engler, follows by arguing that while extraterritorial impact will be real it results are expected to be uneven across sectors and applications, with transparency requirements assessed as producing only minor behavioural changes. The most outright prohibitions are assessed as likely to remain largely confined to the EU market.
Greenleaf (2024) focuses more on the specificity of which channels would lead to a Brussels effect, identifying four distinct mechanisms of EU regulatory influence: direct extraterritorial application, voluntary corporate adoption, legislative emulation by third countries, and incorporation into international standards. He argues that all four must be assessed together to accurately measure the EU's reach.
Czerniawski (2024) looks on the other hand at GDPR enforcement records as a precedent, highlighting that both the GDPR and the EU AI act rely on a destination approach to territorial jurisdiction (meaning that a law applies based on where the effect or output land). He finds that enforcement outside Member States has been persistently difficult under the GDPR, and argues that the AI Act faces comparable constraints.
Marco Almada and Anca Radu (2024) look beyond the more technical distinction, and while focussing on the broader spirit of the AI Act argue that its reliance on a product-safety framework, rather than a fundamental rights framework, may lead to a regulatory model that once diffused globally offers weaker protections than EU policy objectives require.
Across these five papers, the evidence indicates that the AI Act's extraterritorial reach is operational in legal terms but variable in practical effect. Enforcement capacity and sectoral context will thus shape how far the Act's extraterritorial reach translates into practical compliance obligations for non-EU providers.