Abstract

Article 57 of the AI Act requires all European Union (EU) countries to set up regulatory sandboxes by August 2026. While regulatory sandboxes have been explored in existing literature, their specific application within the AI Act context remains understudied, with stakeholders’ expectations largely unexamined. Our research involved interviewing various stakeholders, revealing diverse perspectives that highlight confusion about these upcoming regulatory mechanisms. This study offers two significant contributions: theoretically, we demonstrate the value of applying sociotechnical imaginaries as a framework in an emerging regulatory environment. As a practical contribution, we identify a substantial gap between regulators’ and other stakeholders’ conceptions of AI regulatory sandboxes. Through our analysis, we identified three central themes within stakeholders' imaginaries—(Non-)understanding, Concerns, and Perspectives—which require careful consideration for effective sandbox design and implementation. These findings provide valuable insights for practitioners preparing for upcoming sandboxes across EU member states and offer theoretical implications for conceptualizing regulatory sandboxes, underscoring the importance of studying stakeholders’ imaginaries early in the regulatory development process.

Download

Share

COinS