What is a regulatory sandbox?
A regulatory sandbox is a supervised program in which a regulator lets developers test a new product or technology. Testing runs for a limited time under agreed conditions. The aim is for both sides to learn how the rules apply before the product is widely used.
Also known as: AI regulatory sandbox
Researched and fact-checked by AI, with no human review. 6 sources listed below. How we verify
Last updated
How does a regulatory sandbox work?
In a regulatory sandbox, a regulator and a developer try out something new in a supervised setting. The UK's Financial Conduct Authority (FCA) has its own Regulatory Sandbox. It says the sandbox lets firms from all sectors of financial services test products and services in a controlled environment. The FCA also says the sandbox is not exempt from regulation. A firm carrying out a regulated activity must be authorized or registered first, it says. Its tests are typically small in scale, limited in duration and involve a limited number of consumers. The tools it lists include restricted authorization, individual guidance and, in some cases, waivers or modifications of its own rules.
The European Union has written the idea into law for artificial intelligence (AI). The rule is Article 57 of the AI Act, in the consolidated text dated July 27, 2026. It requires each member state to have at least one AI regulatory sandbox operational at national level by August 2, 2027. The article describes a controlled environment for developing, training, testing and validating innovative AI systems. It says this happens for a limited time before the systems are placed on the market. It says the work follows a plan agreed with the authorities. Participants remain liable for damage caused to third parties.
How are sandboxes used for medical AI?
The UK's Medicines and Healthcare products Regulatory Agency (MHRA) launched AI Airlock on May 9, 2024. It is a regulatory sandbox for AI as a medical device. The agency's program page says two testing phases have been completed with 11 innovators. It says the resulting reports are not formal guidance. A third phase, focused on post-market surveillance, opened for applications on October 6, 2026. Its application page describes three settings:
- roundtable simulation workshops
- virtual testing with real or synthetic data
- real-world use in which the tool's outputs are not used to make decisions about patients
On the same day, the UK government said the MHRA will explore options to introduce staged authorization pathways. It said the options include regulatory sandboxes and support for proposed legislative reforms. Those reforms would create statutory mechanisms for sandboxes, it said.
Sources
- Regulatory Sandbox, Financial Conduct Authority
- Article 57: AI regulatory sandboxes, European Commission, AI Act Service Desk
- MHRA launches AI Airlock to address challenges for regulating medical devices that use Artificial Intelligence, Medicines and Healthcare products Regulatory Agency (GOV.UK)
- AI Airlock: a regulatory sandbox for AI enabled Medical Devices, Medicines and Healthcare products Regulatory Agency (GOV.UK)
- AI Airlock Phase 3 application, Medicines and Healthcare products Regulatory Agency (GOV.UK)
- Government Response to the National Commission's Recommendations on the Regulation of AI in Healthcare., Medicines and Healthcare products Regulatory Agency (GOV.UK)
