Will AI Regulatory Sandboxes Work?

The EU introduced AI regulatory sandboxes under the AI Act of August 2024 to enable companies to experiment with AI systems under regulatory supervision, balancing innovation and safety.
Proponents see the UK’s fintech sandbox as evidence that the model works and that it encourages innovation, regulatory knowledge, and provides a more adaptable regulatory framework.
Critics raise concerns about the effectiveness of sandboxes in achieving their intended goals, systemic risks, and whether smaller firms can adequately participate given regulatory burdens, all pointing to potential inequalities in innovation capacity.
The draft implementing law sets out operational frameworks but introduces uncertainty with an industry-pressured pause and ongoing debates over the extent of regulatory simplification.
The EU Digital Omnibus proposal includes an amendment to establish an EU-level sandbox alongside national ones, aiming at reducing regulatory burdens.
The success of EU’s AI sandboxes hinges on how unresolved design issues are addressed by August 2026, determining if they achieve their dual promise of innovation and regulatory compliance.