GenAI sandbox Singapore: what it means for the legal sector

A managing partner scrolling LinkedIn in August 2026 keeps running into the same term: GenAI Sandbox. It sounds like a natural home for a document review build, and it sounds like it might be cheaper than a consultant’s proposal. Neither assumption survives a look at what the scheme actually funds, who it was built for, and what remains unpublished about how, or whether, the legal sector fits.

The short answer. The GenAI Sandbox is a national programme run by IMDA and EnterpriseSG that lets an eligible Singapore SME trial one pre-approved generative AI solution from a fixed shortlist, with up to 50% funding support for a three-month pilot, per IMDA’s GenAI Sandbox 2.0 factsheet, launched 2 December 2024. It was built for general SME adoption of an existing tool, not for the legal sector specifically, and as of August 2026, neither IMDA nor EnterpriseSG has published a current intake window or a legal-sector solution on the shortlist. A document review or due diligence system built around one firm’s own checklists is a custom project, and that is what the Enterprise Development Grant funds, at up to 50% support for a qualifying SME, verified July 2026 on enterprisesg.gov.sg.

This guide covers what the sandbox has actually funded so far, why its shape does not match a law firm’s document review workflow, where the real legal-sector programmes sit instead, and what a firm scoping a project now should do while the sandbox’s current status stays unpublished.

What the GenAI sandbox actually funds for Singapore’s legal sector

The sandbox is a trial, not a build. IMDA picks a shortlist of GenAI solutions, an eligible SME picks one from that shortlist, and the pilot runs for three months with up to 50% of the cost covered, with a further nine months of funding support available to a firm that continues on the same tool past the pilot, per the GenAI Sandbox 2.0 factsheet. None of that involves the firm’s own document templates, clause libraries, or review checklists, because the tool on offer is whatever IMDA has already vetted and listed.

The first cohort ran from February to May 2024.

The second cohort, GenAI Sandbox 2.0, launched on 2 December 2024 with roughly 15 solutions across three categories, targeting more than 300 SMEs. Both cohorts describe their target sectors as retail, food and beverage, education, and hospitality, among others. IMDA’s public materials do not list a legal-sector or professional-services solution on either shortlist, and Kept has not found a published 2026 intake window as of August 2026. That is the sandbox’s current status: confirmed as a real, running programme, unconfirmed as a fit for a Singapore law firm. Confirm current sector scope and any open intake directly on imda.gov.sg before an application goes anywhere near a partner’s desk.

Eligibility for the sandbox mirrors the eligibility test a law firm already knows from the EDG: ACRA-registered in Singapore, operating here, and group turnover of S$100 million or less a year, per EnterpriseSG and IMDA’s original launch materials, February 2024. Nearly every Singapore law firm clears that bar on paper. Clearing the SME test is not the open question, though. The open question is whether a solution the firm actually needs, a system that reads a data room the way one practice group works it, sits on a shortlist IMDA controls, and nothing published says it does.

Two programmes share the word “sandbox”

Part of the confusion in the search term itself is that Singapore runs more than one government initiative with “sandbox” in the name, and they do different jobs. The GenAI Sandbox for SMEs, covered above, is a funding and trial programme for adopting an existing tool. A separate initiative, the Global AI Assurance Sandbox, run by IMDA and the AI Verify Foundation, is a testing ground where builders and deployers of GenAI applications get their systems evaluated by specialist technical testers. It funds nothing for the SME running the trial. It exists to test AI systems against emerging assurance standards, not to subsidise buying one. A firm that hears “sandbox” in a legal-tech pitch should ask which of the two is meant, because the funding one and the testing one answer completely different questions.

Where the legal sector’s real 2026 programmes sit instead

The legal sector already has its own framework, and it is not the sandbox. Singapore’s Ministry of Law published the Guide for Using Generative AI in the Legal Sector on 6 March 2026, setting out the compliance principles a firm’s AI use has to sit inside regardless of which grant pays for the build. Alongside the guide, MinLaw runs the LIFT initiative, a firm-facing tech adoption pilot that succeeded the earlier Tech-celerate for Law programme. LIFT’s pilot is scheduled to end in the third quarter of 2026, and Kept has not independently verified its scope, funding mechanics, or successor programme against an official source, so none of that detail appears here as a figure to plan around. A firm that hears “sandbox” in a legal-tech conversation is more often hearing about LIFT, the MinLaw guide, or a vendor’s own demo environment than about IMDA’s SME programme, and the three are not interchangeable.

Why a document review build is an EDG project, not a sandbox trial

A first-pass document review system worth building reads the deal type a specific practice group handles, flags what changed against the firm’s own prior draft, and drafts the first cut of a review memo in the firm’s own format. None of that comes off a shortlist. Set the two funding shapes side by side and the mismatch is structural, not a matter of degree.

The training row is the one worth sitting with. A trial of someone else’s pre-approved tool ends after three months (or twelve, with the extension) with the firm still dependent on that vendor for every update. A custom EDG build ends with the firm’s own associates able to run and correct the system without the consultant in the room, which is the difference between renting a capability and keeping one. The wider argument for why that distinction determines whether an AI system survives past its first year, rather than quietly reverting to the old spreadsheet six months after go-live, is in why AI projects fail.

The oversight requirement runs the same way under either funding route. MinLaw’s guide asks that a lawyer checks the output before it reaches a client, whether the system came from a sandbox trial, an EDG-funded build, or a subscription bought outright. A three-month pilot of a shortlisted tool does not exempt the firm from that principle, and neither does a custom build. The grant changes who pays for the system. It does not change who signs off on what it produces.

Worked example: the EDG path, priced today

The sandbox cannot produce a worked example for a law firm, because no legal-sector solution or rate is published. The EDG can, because its figures are verified.

Applications go through the Business Grants Portal before a vendor is paid anything, and processing runs 8 to 12 weeks. A firm that can name the workflow and the baseline hours today has a fundable application well before any sandbox intake question gets answered. The full mechanics, including the consultant certification rule that trips up legal-sector scopes, are in our EDG grant guide for Singapore law firms.

Run the same S$40,000 build through the sandbox’s structure and the numbers stop working before the arithmetic starts. Even setting the unpublished shortlist aside, the sandbox funds a three-month pilot of one vendor’s existing product. A custom S$40,000 build is not a three-month pilot of an existing product. It is a project, scoped around one firm’s workflow, of the kind the EDG was built to fund and the sandbox was not.

What to watch

Four things are still open, and a firm should treat each as a reason to confirm rather than assume.

Whether the sandbox adds a legal or professional-services solution to a future shortlist is unpublished. Whether a 2026 cohort is currently open for applications is unpublished. LIFT’s status past its Q3 2026 pilot end date is unpublished. And EnterpriseSG’s EDGE grant, which consolidates the EDG, PSG, and Market Readiness Assistance into a single scheme in the second half of 2026, has not published the support percentages that will replace the EDG’s verified 50%. Our EDGE grant guide covers what is confirmed there and the filing window that still favours moving on the EDG now rather than waiting.

None of these four unknowns changes what is already true. A Singapore law firm running first-pass document review by hand has a cost, measurable in hours per file, and a funded path to fix it that does not depend on any of them resolving.