AI regulatory sandbox
In plain wordsA supervised testing ground where companies can try out new AI with regulators watching and advising, before the product faces the full weight of the rules.
A time-limited, supervised space set up by a public authority in which a provider can develop and test a new AI systemAI system, n.: A complete piece of technology that uses AI to do something, such as screening loan applications, driving a car or chatting with customers, including the AI model and the software around it., in real-world conditions where appropriate, under a plan agreed with the regulator. Under the EU AI ActEU AI Act, prop. n.: The European Union's law on artificial intelligence, the first broad AI law anywhere: it bans some uses, sets strict checks for risky ones and puts duties on the makers of the biggest AI models., every member state must have at least one national AI regulatory sandbox operational by 2 August 2027, and the AI OfficeAI Office, prop. n.: The European Commission's team that oversees the EU AI Act, especially the rules for the largest AI models, and coordinates how the law is applied across Europe. may run an EU-level sandbox for systems it supervises. Participants receive compliance guidance; small companies get priority and free access; and exit documentation can support conformity assessmentconformity assessment, n.: The check, done by the maker or by an approved outside body, that a high-risk AI system meets the EU's legal requirements before it can be sold or used.. Participants remain liable for harm to third parties, but authorities do not fine those that follow the plan and guidance in good faith. Personal data collected for other purposes may be reused there to develop certain public-interest systems under strict conditions.
The start-up tested its triage model in the national AI regulatory sandbox before seeking CE marking.
Origin
The regulatory sandbox comes from financial regulation: the UK Financial Conduct Authority opened the first fintech sandbox in 2016, and the model spread to data protection and other fields. Spain launched a pilot AI sandbox with the European Commission in 2022; the AI Act made national sandboxes mandatory, and the 2026 Digital OmnibusDigital Omnibus, prop. n.: A European Union package, proposed in 2025, that simplified and delayed parts of its digital rules — above all pushing back the strictest requirements of the EU AI Act by more than a year. moved the deadline from August 2026 to August 2027.
Usage note
Unrelated to a sandboxsandbox, n.: A sealed-off computer space where an AI agent can run programs and change files without being able to harm the real system, reach private data or contact the outside world unless allowed. in software and agent security, which isolates running code. Critics worry that sandboxes can become channels for regulatory favoritism; supporters see them as one answer to the Collingridge dilemmaCollingridge dilemma, n.: The trap that a new technology is easiest to control when little is known about its harms, but by the time the harms are clear it is too widespread to change easily..
Related
- EU AI ActEU AI Act, prop. n.: The European Union's law on artificial intelligence, the first broad AI law anywhere: it bans some uses, sets strict checks for risky ones and puts duties on the makers of the biggest AI models.
- Collingridge dilemmaCollingridge dilemma, n.: The trap that a new technology is easiest to control when little is known about its harms, but by the time the harms are clear it is too widespread to change easily.
- conformity assessmentconformity assessment, n.: The check, done by the maker or by an approved outside body, that a high-risk AI system meets the EU's legal requirements before it can be sold or used.
- AI OfficeAI Office, prop. n.: The European Commission's team that oversees the EU AI Act, especially the rules for the largest AI models, and coordinates how the law is applied across Europe.
- risk-based approachrisk-based approach, n.: A way of regulating AI in which the strictness of the rules depends on how much harm a use could cause: bans for the worst uses, heavy checks for risky ones, little or nothing for the rest.
- sandboxsandbox, n.: A sealed-off computer space where an AI agent can run programs and change files without being able to harm the real system, reach private data or contact the outside world unless allowed.
Defined in law or standards
EU AI Act (Reg. (EU) 2024/1689), Art. 3(55), Arts. 57–59; deadline as amended by Reg. (EU) 2026/1744
Mentioned in
Digital OmnibusDigital Omnibus, prop. n.: A European Union package, proposed in 2025, that simplified and delayed parts of its digital rules — above all pushing back the strictest requirements of the EU AI Act by more than a year., sandboxsandbox, n.: A sealed-off computer space where an AI agent can run programs and change files without being able to harm the real system, reach private data or contact the outside world unless allowed., Collingridge dilemmaCollingridge dilemma, n.: The trap that a new technology is easiest to control when little is known about its harms, but by the time the harms are clear it is too widespread to change easily., Texas Responsible Artificial Intelligence Governance ActTexas Responsible Artificial Intelligence Governance Act, prop. n.: A Texas law, in force since 2026, that bans a short list of harmful uses of AI — such as pushing people toward self-harm or making sexual deepfakes of children — and sets rules for government use of AI.