Regulatory Sandboxes: One Decade On
Regulatory sandboxes have spread like wildfire since the U.K. Financial Conduct Authority launched its sandbox for financial technology businesses (fintech) one decade ago. Despite widespread adoption, however, there is little empirical evidence available to assess whether the signature sandbox policy combination of regulatory rollbacks and regulatory guidance is in fact good policy. The empirical evidence that is available suggests that regulatory sandboxes are beneficial for the tech firms that participate in them, but tells us nothing about how regulatory sandboxes have impacted the broader enterprise of regulation, or whether the innovation generated by sandbox participants is beneficial for anyone other than the innovating firms themselves. From the outset, there were reasons to be concerned about sandboxes’ deregulatory impact and grounds for skepticism about the type of regulatory learning that sandboxes would facilitate. A decade of experience with fintech sandboxes has not allayed those concerns; sometimes it has deepened them.
Careful attention to sandbox design features can mitigate some concerns, but we should not skip straight to design questions without first considering whether a regulatory sandbox is appropriate at all. A reckoning with the sandbox model is particularly necessary at this moment, when there is a push to use sandboxes to further innovation in generative artificial intelligence (AI). It is becoming increasingly clear that simply scaling up generative AI will not solve its limitations, just as it is becoming increasingly clear that generative AI tools
have significant negative impacts on privacy, intellectual property rights, and our environment (among other things). In these circumstances, it seems foolhardy to rush headlong into adopting sandboxes that roll back legal protections in order to let AI thrive.