UK AI Growth Lab Regulatory Sandbox - Consultation on Two Models
UK-AIGLS · UK Department for Science, Innovation and Technology
The UK Department for Science, Innovation and Technology launched a consultation in October 2025 on the AI Growth Lab, a proposed regulatory sandbox enabling companies to test AI innovations under modified regulatory conditions. Two structural models are under consideration: a centrally operated sandbox administered by the government across multiple sectors, and a regulator-operated model in which a designated lead regulator manages each sandbox instance. The initiative is intended to reduce compliance barriers for AI development while maintaining appropriate oversight.
Overview
The AI Growth Lab consultation, opened in October 2025 by the UK Department for Science, Innovation and Technology (DSIT), sets out two competing governance architectures for a national AI regulatory sandbox. Under the central government model, DSIT or a designated central body would coordinate sandbox operations across sectors, providing a single point of entry for applicants. Under the regulator-led model, a sector-specific lead regulator would be appointed for each sandbox instance, bringing domain expertise to bear on the regulatory modifications granted. Participants admitted to either model would operate under time-limited, targeted relaxations or adaptations of existing regulatory requirements, allowing real-world testing of AI systems that would otherwise face uncertain or prohibitive compliance obligations. The consultation seeks stakeholder input on criteria for admission, duration of sandbox participation, safeguards for affected third parties, and how learnings will inform future regulatory design. No binding obligations attach to enterprises during the consultation phase, but the eventual framework will directly affect compliance planning for AI developers and deployers seeking to operate in the UK market.
Key Requirements
- •Organisations wishing to participate must apply and meet admission criteria to be defined following consultation closure.
- •Sandbox participants will operate under targeted, time-limited regulatory modifications specific to the AI system under test.
- •Two governance models are under consideration: centrally operated by government or led by a designated sector regulator per sandbox instance.
- •Safeguards for consumers and third parties affected by sandbox-tested AI systems must be maintained throughout participation.
- •Learnings and outcomes from sandbox activity are expected to be shared with regulators to inform future AI regulatory policy.
- •Consultation stakeholder responses are due to DSIT; final model selection and implementation timeline have not yet been published.