This guide is written for solicitors, trainees and law students in England and Wales.
An HMCTS data licence is what allows you to run computational analysis on published court and tribunal lists. Without one you have no permission to copy, adapt or redistribute that data at all, because the Crown holds database rights in it. This checklist covers which of the two licences to apply for, what the application needs, and what happens after the Governance Panel decides.
Who needs one?
Anyone performing computational analysis on the lists published through the Court and Tribunal Hearings Service. In practice that is lawtech suppliers building listing, analytics or monitoring products, firms building the same tools internally, and anyone whose research design involves systematic collection rather than reading a list.
The threshold is the method rather than the purpose. Reading a cause list is not caught. Scraping it, storing it and running analysis over it is.
It is worth being clear about what the licence does not reach, because the two are easily conflated. This is a licence over the listing data published through the hearings service, not over judgments, which are published under their own arrangements through Find Case Law and carry a separate licensing position. A product that reads judgments and a product that reads cause lists are subject to different regimes, and a supplier building both needs to have worked out which parts of its pipeline sit where before it applies for anything.
Which licence should you apply for?
Two are available, and the choice is largely made for you by what the product does.
| Experimental | Transactional | |
|---|---|---|
| Term | 1 year | 3 years |
| Permitted use | Proof of concept only | Commercial products |
| Release to customers | Not permitted | Permitted |
| Release to the public | Not permitted | Permitted |
| Available for high-risk uses | Yes, and only this | No |
| Cost on initial grant | Free | Free |
Both are free when first granted, with HMCTS reserving the right to charge later.
The practical difficulty with the Experimental licence is not its restrictions but its arithmetic. Twelve months is a short runway for anything that has to be built, tested against real listing data and then demonstrated to a panel, and a supplier that spends the first three months on procurement and data engineering has spent a quarter of the term before producing anything a reviewer could look at. Plan the demonstration first and build backwards from it, rather than treating the panel review as something that happens after the product is finished.
Which uses are treated as high risk?
Four, and each is capped at an Experimental licence:
- Fully automated legal advice.
- Anticipating legal decisions for a client or consumer through predictive analytics or modelling.
- Otherwise directly informing whether a third party pursues legal action.
- Personal profiling, particularly where court data is combined with other datasets.
Read together, those four cover a good deal of what litigation analytics is marketed as doing. A tool that scores a claim’s prospects, or that tells a client how a judge has decided comparable applications, falls inside the second or third on any sensible reading. That does not make it unlawful. It does mean the route to market runs through a one-year licence with no commercial release, then through a renewal decision nobody has yet made. We covered what the restriction means for the analytics market when the licence was published.
What the application needs
Work through these before starting the form rather than during it.
- A clear statement of the intended use, written against the four high-risk categories rather than around them. An application that describes an outcome-prediction tool in neutral language still describes an outcome-prediction tool.
- A data protection impact assessment. Required, and the one document most likely to be thin. Court listings carry names.
- An algorithmic transparency recording standard template, where relevant to what you are building.
- Your position against the twelve assessment principles, which cover open justice, judicial independence, protection against bias and algorithmic transparency among others.
- Retention periods, stated rather than implied.
- Named accountability for the analysis and for the data held.
What happens after you apply?
The Governance Panel is assessing a proposition rather than a piece of software, which is why applications that read as engineering documents tend to do badly. What the twelve principles are aimed at is the effect of the analysis on the administration of justice, so an application that explains what the tool does without explaining what it does to open justice, judicial independence or the risk of bias has answered a question nobody asked.
Applications go to a Governance Panel, which assesses them against the twelve principles. Two features of the process deserve more attention than they have had.
The first is publication. HMCTS will publish both successful and rejected applicants, with reasons. A refusal therefore becomes a public document naming a business and describing its intended product, which changes the calculation considerably: an application is not a low-cost enquiry that can be withdrawn quietly if it goes badly.
The second is timing. Appeals must be brought within 20 working days of the decision. An Experimental licence runs 12 months against a Transactional licence’s 36, so a proof of concept that takes nine months to build leaves three in which to demonstrate anything to the panel that granted it. Nothing in the guidance promises a renewal will follow.
What should a buying firm ask?
The licence is as useful to a firm procuring analytics as to the supplier building it, because it gives a purchaser something concrete to ask about in place of the general assurances that usually fill this part of a diligence conversation. A supplier holding a Transactional licence has been through a panel assessment against twelve published principles, which is a meaningful credential; a supplier holding an Experimental one cannot lawfully release the product to you at all, whatever the sales conversation suggests.
Four questions get you most of the way. Which licence do you hold, and when does it expire? Which of the four high-risk categories does your roadmap touch, now or in the next version? Was your application published, and what did the reasons say? And if the product ingests court listings alongside other datasets, how is the personal profiling risk handled? None of these requires technical knowledge to ask, and a supplier that cannot answer them quickly has probably not applied.
Common mistakes
Assuming publication means the data is free to use. Published is not the same as licensed. The Crown’s database right in the lists is what the licence addresses.
Treating the DPIA as a formality. It is the document the panel will read most closely, because the personal profiling category sits directly on top of it.
Building first and applying later. A finished product that turns out to sit in a high-risk category has been built against a licence that will not permit its release.
Confusing this with the AI sandbox. The Advisory AI Growth Lab gives coordinated regulator access and confers no permission over anything. The data licence is the permission. They are separate processes with separate deadlines.
Official sources
The application guidance sets out each stage and is the document to work from. The licence publication page carries the licence terms themselves, and the Court and Tribunal Hearings Service is the source the licence governs.
Without a licence you may not copy, adapt or redistribute the data at all. With the wrong one you may build but not sell. The four high-risk categories are drawn widely enough to capture most outcome-prediction tools, so the licence question belongs at the start of a product plan rather than at the compliance review before launch.
Last updated: 6 August 2026.