We then identify governance challenges across three sandbox phases—pre-testing, testing, and post-testing—and propose structured solutions. 5/

Our article examines the AI Act’s sandbox framework, arguing it shifts from traditional models by emphasising regulatory learning over mere tech disruption, and by broadening public interest goals to include strategically aligned commercial innovations. 4/

Regulatory sandboxes can offer a practical way to address this dilemma by enabling controlled, evidence-based testing of new technologies. 3/

Collingridge called the “dilemma of control” the challenge of regulating emerging technologies: act too early and risk stifling innovation; act too late and potential harms proliferate unchecked. 2/

We found that while many DPAs have some hands-on experience with AI cases, their activity levels vary widely. Still, their growing expertise could play a key role in enforcement, especially if supported by supranational coordination to help the less active authorities.

How ready are DPAs to enforce the AI Act? To find out, we looked at what they’re already doing: their decisions and activities related to AI and automated decision-making across all EU Member States. Thus, we created an index to measure how active they are in this area. 2/