8 min read · AI regulation · UAE
What is the UAE AI Sandbox?
Capio team

Dubai’s AI Sandbox is a government run programme for real world testing of AI with regulators. See who can apply, what support you get, how long it runs and how it differs from national sandboxes and hackathons.
The UAE AI Sandbox most people are searching for is Dubai’s AI Sandbox. It is a government run programme operated by Sandbox Dubai under Dubai Future Foundation. It enables companies to test AI in the real world with the right regulators and government entities involved. You apply during an open call and, if selected, you design a monitored experiment with defined safeguards and measures. Official programme page
Dubai Future Foundation announced the AI Sandbox on 28 October 2026, with applications open until 18 December 2026. Check the official materials for any changes to dates. Launch and deadline
Is it the same as Dubai’s AI Sandbox?
Yes. When people say “UAE AI Sandbox” they usually mean Dubai’s cross sector AI Sandbox run through Sandbox Dubai. The platform sits within Dubai’s broader regulatory modernisation push and coordinates with the relevant Dubai and federal regulators for each use case. Sandbox Dubai Executive Council approval
Who runs the AI Sandbox and what is the mandate?
Sandbox Dubai runs the programme under Dubai Future Foundation. The mandate is to enable cutting edge innovation under adaptive regulatory supervision. That means structured experiments in live or live like settings, with regulators engaged throughout to observe, learn and shape next steps. Programme overview
Who can apply and when do applications close?
An open call invites applications. For the current cohort the window runs to 18 December 2026. Use the official portal to apply and to confirm dates. Open call announcement AI Sandbox page
What level of product maturity do you need?
- The public materials do not set a fixed level, but the focus is real world testing. In practice, teams should expect to present a working product and a specific regulatory or policy challenge to test.
What sectors and enabling technologies are in scope?
- Priority areas include education, tourism and hospitality, transport and mobility, logistics, media and government services. Enabling technologies like foundation models are also welcomed. Sectors in scope
What support do selected participants actually get?
Selected teams co design an experiment with Sandbox Dubai and the relevant regulators. Support typically includes:
- Real world testing in a controlled setting, with monitoring
- Facilitation of data access where appropriate, subject to the data owner’s approval. Access is not guaranteed
- Ongoing engagement with regulators and, where appropriate, regulatory flexibility limited to the experiment
- Technical and commercial guidance tailored to the use case
Sources: AI Sandbox page How the process works

How long does a sandbox experiment last?
Expect a multi month run. Sandbox Dubai describes a typical testing period of six to eighteen months, with the FAQ noting that six to twelve months is common. Duration varies by use case and the controls required. Process and durations FAQ
What happens after testing and how are next steps decided?
- After monitoring and evidence collection, the programme runs a final evaluation with the participants and regulators. That readout informs what comes next, for example further piloting under standard permissions or work on updated guidance. Final evaluation step

What the AI Sandbox is not
Does acceptance grant automatic regulatory approval?
- No. Selection to the cohort does not equal permission to launch commercially. Permissions and any flexibility are limited to the experiment and subject to permit issuance. Permit issuance in process
Is access to government or private data guaranteed?
- No. Sandbox Dubai may help facilitate access, but approval sits with the data owner and may not be granted. Data access note
How does Dubai’s AI Sandbox differ from national UAE sandboxes?
- TDRA ICT Regulatory Sandbox: a national sandbox for ICT services and solutions. It provides an agile space to test under TDRA oversight. Use this when your test sits squarely in telecoms and ICT regulation at federal level. TDRA ICT sandbox
- Central Bank sandbox: a national framework for fintech experimentation under the Central Bank of the UAE. Use this for financial services innovations in scope of CBUAE rules. CBUAE sandbox conditions
- UAE Regulations Lab (RegLab): a federal lab established in 2019 to enable testing under time bound licences and to shape federal regulations for emerging tech, including AI. RegLab overview
Dubai’s AI Sandbox is an emirate led, cross sector programme that convenes the right Dubai and federal entities around a specific AI use case. The national sandboxes above are domain specific and run by federal bodies.

How do you avoid confusion with non government “sandbox” uses?
The word “sandbox” is also used for events and lab environments that are not regulatory sandboxes. For example, the ITU’s AI for Good Lab is running a UAE hackathon on 8 December 2026 with mentoring, cloud credits and CHF prizes. That event does not grant any regulatory permissions in the UAE. ITU AI for Good Lab hackathon
What evidence should your application include?
You improve your chances with clear, testable intent and safeguards. Strong applications usually show:
- A precise problem statement and the regulatory or policy barrier you need to test
- A working product, a narrow experiment scope and a plan for how you will measure safety and performance
- Evidence of user or market demand, and why testing in Dubai is relevant
- Specific success metrics for the test period, and how you will collect and share them with regulators
What documentation should you prepare before you apply?
Have practical artefacts ready to support scoping and permit discussions:
- A high level data flow map for the experiment, including data sources, transfers, storage and deletion
- Responsible AI guardrails for the use case, including when a human must review or override
- Impact and risk assessments suited to the use case, with mitigations
- Security, auditability and logging plans, especially for agent based systems that act on user data
- A clear plan for testing datasets and any model updates during the run

What does success look like inside the sandbox?
Success is evidence. A good run leaves you and the regulators with:
- Measured safety and performance against the agreed success metrics
- A record of risks encountered and how your controls worked in practice
- Clarity on the path forward, whether that is a standard permission route, further trials or updated guidance
Common reasons applications fail
- The regulatory or policy question is unclear or too broad
- The product is too early for real world testing
- Risk controls and monitoring are thin
- The plan assumes data access that is not within the programme’s gift to grant

Frequently asked questions
Can international companies apply?
The open call is public and the programme page describes who can apply. It does not state nationality limits on the public page. Confirm eligibility and any local setup needs in the application portal. AI Sandbox page
Do you need to be established in Dubai already?
The public materials do not say you must already be incorporated in Dubai to apply. Depending on the test, sector approvals or permits may still be required. Use screening to confirm what is needed. Process overview
What happens to IP created during testing?
IP terms are not detailed on the public pages. Clarify ownership and usage in your participant agreement, data sharing agreements and experiment plan.
How are permits and exemptions handled?
Permit issuance is a defined step in the process. Any regulatory flexibility applies to the specific experiment and operates under monitoring. It does not equal market authorisation. Permit step
How does CAPIO PRO approach AI sandbox readiness?
We help teams arrive ready with a credible plan, controls and measures that stand up in front of regulators.
-
Readiness assessment and AI Health Score
- What we do: a three to four week assessment across leadership, governance, data, technology, capability, measurement and trust, producing a baseline AI Health Score.
- Why: it surfaces strengths and gaps before you propose an experiment, so your plan is realistic and measurable.
-
Governance foundation matched to UAE regimes
- What we do: set responsible AI guardrails, human in the loop rules, an AI inventory and risk register, and a data residency map held in your AI Twin.
- Why: regulators look for clear accountability and controls you can evidence.
-
Pilots through gates with measured results
- What we do: structure work into a proof of concept, MVP and then an enterprise solution owned by your team, with clear exit gates and metrics.
- Why: it keeps scope tight and evidence rich, which suits monitored sandbox testing.
-
Fractional Chief AI Officer where needed
- What we do: provide leadership for the first ninety days to set the baseline, governance and first decisions, then hand over.
- Why: it aligns product, engineering, legal and compliance around the experiment without adding permanent headcount.
Where to start
If you plan to apply, start by writing the problem you want to test and the evidence you will bring back. If you want a partner to pressure test your plan and get you experiment ready, get in touch.


