AI regulatory sandbox
A controlled space where an AI system is developed and tested under the regulator's direct supervision, before any market placement.
The UK's FCA opened the first regulatory sandbox in 2016; Article 57 of Regulation (EU) 2024/1689 requires each Member State to run one for AI, operational by August 2, 2026. A regulatory sandbox is an arrangement in which a company develops and tests a system under the direct supervision of the competent authority, following an agreed plan, with guidance and sometimes temporary accommodations of certain requirements. The mechanism answers a difficulty specific to fast-moving technology: the regulator does not know the product well enough to issue a useful rule, and the company does not know the rule's interpretation well enough to invest, so each waits for the other. The sandbox breaks that standoff by producing doctrine from concrete cases. The European artificial intelligence regulation makes establishing one mandatory for every member state, with priority access for small companies, precisely because compliance cost is the main obstacle they face. Two caveats come with the tool. Taking part never exempts anyone from liability for damage caused. And selective access can hand a competitive edge to those admitted, which is why selection criteria must be public.
Regulation (EU) 2024/1689 on artificial intelligence, published on July 12, 2024 and in force from August 1, 2024, requires each member state to establish at least one operational regulatory sandbox by August 2, 2026, with priority access for small and medium-sized enterprises.
Follow the research
One letter a week. First, the week's underwriting note, when one is published: a dated, argued decision, signed by Paul Bernardski. Then what the observatory published on cyber risk, political risk and insurance. No account, no tracking.
No spam, one-click unsubscribe, no data ever sold.