The UK government is launching AI Growth Labs: regulatory sandboxes where organisations test AI applications in a secure environment and work through the regulatory questions directly with the regulators concerned. Legal services goes first — applications open later this summer — with other sectors to follow later in 2026, according to Osborne Clarke's June regulatory outlook.
Sandboxes are easy to dismiss as innovation theatre. This one deserves more attention, for a structural reason: the UK has chosen to regulate AI through existing sector regulators rather than a single statute, and a sandbox is what makes that model usable. When no act of Parliament defines what is permitted, the only way to resolve uncertainty is to ask the regulator — and the Growth Labs institutionalise the asking.
What Was Actually Announced
- AI Growth Labs. Regulatory sandboxes for testing AI applications, described as "a safe space in which to test innovative legal services products and discuss regulatory issues directly with regulators". Legal services first; applications open in summer 2026; rollout to further sectors later in the year.
- Cross-regulator structure. For the legal lab, the Information Commissioner's Office is collaborating with the Solicitors Regulation Authority, the Council for Licensed Conveyancers, and the Legal Services Board on cross-regulatory challenges — the data protection regulator and the sector regulators in one room.
- The wider June package. The same period brought the AI Hardware Plan (a strategy for developing and scaling the semiconductor technologies underpinning AI), a new AI Economics Institute under HM Treasury and DSIT to assess AI's impact on productivity and labour markets, and two DRCF calls for input — one on consumer attitudes to generative and agentic AI, one on tools for managing AI risks, open until 2 September 2026.
A June 2026 House of Commons Library briefing confirms the frame all of this sits in: there is no UK AI Act and no AI bill currently before Parliament. The sector-based approach is not a gap waiting for legislation — it is the policy.
Who This Affects, and How
Legal services firms and legal tech vendors have a decision with a deadline. If you are building or buying AI that sits near the boundaries of regulated legal activity — automated advice, AI-assisted conveyancing, agentic drafting that touches client files — a sandbox place converts regulatory uncertainty into documented dialogue with the SRA, the CLC, and the ICO at once. That documentation has value beyond the lab: it is the answer to the professional indemnity insurer and the risk committee.
Everyone else should read the legal lab as a template. The cross-regulator structure being tested — data protection plus sector regulator, jointly — is the obvious model for the sectors that follow, and the government has said rollout continues later this year. The practical move now is to know which of your planned AI deployments are stuck on regulatory uncertainty, so that when your sector's lab opens you have a candidate ready rather than a brainstorm to run.
Compliance and risk teams get a second, quieter opportunity: the DRCF's call for input on AI risk-management tools is open until 2 September. Regulators are asking what frameworks organisations actually use to manage generative and agentic AI risk. If your answer is well-developed, contributing shapes the guidance you will later be assessed against — a point we made in our analysis of the ICO's direction of travel, where waiting for final guidance has repeatedly meant inheriting rules written around other people's practices.
What Participation Actually Looks Like
A sandbox place is not a regulatory holiday. Nothing announced suspends UK GDPR, professional conduct rules, or consumer law. What it changes is the order of operations: instead of building, launching, and discovering the regulator's view at enforcement time, you test with the regulator's view in the room. For AI products where the compliance question is genuinely novel — who is responsible when an agentic system gives reserved legal advice? — that reordering is worth more than any written guidance, because the guidance does not exist yet.
The cost is disclosure. Sandbox participation means showing a regulator your product's failure modes early. Organisations whose AI governance is aspirational rather than operational should treat that as the entry requirement it is: the lab rewards teams who already know how their system behaves, monitor what it does, and can isolate what it touches — the same fundamentals we set out for sandboxing agent workloads at the infrastructure level.
What to Do Before Applications Open
- Inventory the stuck projects. List AI deployments that stalled on regulatory uncertainty rather than technical difficulty. Those are your sandbox candidates — and if there are none, you do not need the lab.
- If you are in legal services, prepare now. Applications open this summer. A credible application will name the regulated activity in question, the specific uncertainty, and the evidence you would generate in the lab.
- Respond to the DRCF by 2 September if your AI risk framework is real. Shaping the assessment criteria beats inheriting them.
- Watch the sector rollout. The government has committed to further sectors in 2026 but not named them. When yours is announced, the legal lab's structure tells you what to expect: your sector regulator plus the ICO, jointly.
The Growth Labs are one piece of a policy estate that now spans compute, hardware, institutions, and regulation — we map the whole of it in our guide to the UK AI policy landscape.
We publish pieces like this regularly — no gate, no sales sequence. Get new posts by email.
Frequently asked questions
What are the UK's AI Growth Labs?
AI Growth Labs are government-backed regulatory sandboxes: secure environments where organisations can test AI applications and discuss regulatory issues directly with the regulators who oversee them, before committing to full deployment. The programme starts with legal services, where the Information Commissioner's Office is collaborating with the Solicitors Regulation Authority, the Council for Licensed Conveyancers, and the Legal Services Board on cross-regulatory challenges. Applications for the legal services lab open in summer 2026, with other sectors to follow later in the year.
Which sectors can apply to the AI Growth Labs?
Legal services is the first sector, with applications opening in summer 2026 for innovators building AI-enabled legal products. The government has said the programme will be rolled out to other sectors later in 2026, though it has not yet named them. Organisations outside legal services should treat the first lab as a template: the cross-regulator structure being tested there — a data protection regulator working alongside sector regulators — is the model likely to be replicated.
What do AI Growth Lab participants actually get?
A safe space to test innovative products and direct access to regulators to work through compliance questions before launch. For AI products whose regulatory treatment is unclear — automated legal advice, AI-assisted conveyancing, agentic workflows that touch client data — that direct channel is the substantive benefit: it converts regulatory uncertainty, which boards read as risk, into documented dialogue with the bodies that would otherwise enforce against you.
How do the AI Growth Labs fit into UK AI regulation?
They are the enabling half of the UK's sector-based approach. The UK has no single AI Act — a June 2026 House of Commons Library briefing confirms regulation continues to run through existing sector regulators. The Growth Labs give that model a front door: rather than a statute defining what is permitted, innovators test propositions with the relevant regulators directly. The same month brought the AI Hardware Plan, a new AI Economics Institute, and two DRCF consultations on AI risk — the sector-based model is being built out, not replaced.
← All posts