A regulatory sandbox is a framework in which a regulator allows companies to test new technologies, products, or business models for a defined period, within set limits, while being exempt from or subject to relaxed versions of certain rules. The regulator supervises the trial, observes risks and benefits, and uses the results to inform future regulation. The term came into wide use after the United Kingdom's Financial Conduct Authority introduced such a scheme for fintech firms in 2015, and many countries have since extended the approach to areas such as finance, healthcare, and mobility.
In AI policy, regulatory sandboxes serve as supervised environments for developing and testing AI systems; the EU AI Act, for example, requires member states to establish AI regulatory sandboxes. The concept is distinct from a sandbox in computer security, which refers to an isolated environment for running programs.