Skip to content
Sign in

Legal

Terms of service

The agreement between your organisation and Dober Security ApS for using Gelex. In short: use it as much as you like for your own work; don't copy the corpus out in bulk, resell it or train a model on it. We show the law and check citations, but we don't give legal advice, run no AI model of our own, and never train any model on your content.

Last changed October 21, 2026
These pages may still change
Gelex has not launched yet. We may still adjust these pages before launch, for example once we have appointed our data protection officer. Every change is dated in the changelog at the foot of the page.

1. Who we are and who these terms are for

Gelex is provided by Dober Security ApS, a Danish private limited company ("we", "us"). Our address, company registration and contact details are in the imprint. These terms apply to every organisation and person that uses the service, on the website, in the app, through the API or through the MCP connector ("you").

If you are a consumer (a private person acting outside your trade or profession), the consumer terms also apply and take precedence where they give you more rights.

2. How the contract is made

  1. You create an account and an organisation. That starts a free trial on these terms; nothing is charged.
  2. To buy a plan, you choose it in Settings → Plan, see its price, period and allowance, and confirm with the order button. Before you confirm you can go back and change any choice; mistakes in your details can be corrected in Settings at any time.
  3. The contract is made when the first payment goes through; we email the invoice as confirmation. Institutions that ask for a quote get a written offer; the contract is made when they accept it.

The contract is in English. We keep the terms in force when you ordered, and you can read, save and print the current version on this page at any time (the date above shows when it last changed). We are not bound by a code of conduct.

3. What the service is

The service is a research corpus of primary legal sources (legislation, case law, tax authority practice and preparatory works) taken from each publisher's own text, structured, graded and indexed by us, and made searchable by you or by an AI model you choose. It also offers matters, document checks, templates and drafting, and signing, as your plan includes them.

We do not supply the AI model. You connect your own model by API key in the app, or use our MCP connector inside an assistant you already use. The model's output is produced by that model and its provider, not by us (see AI output).

4. Accounts, organisations and seats

An account belongs to a person; a subscription belongs to an organisation. The organisation's owners and admins decide who is a member, what each member may do, and which AI connections and API keys exist. You are responsible for what happens under your accounts and keys, and for keeping sign-in methods and keys secret.

A seat is one named person. Sign-in details are personal, also within your organisation: no shared logins. A seat may be moved to someone else when its holder leaves or changes role, not passed round in turn.

5. What you may do

For the term of your subscription we grant your organisation a non-exclusive, non-transferable right for its members to search, read, quote, cite and print material from the service for their own work, including advice to their own clients, and to include quotations with their citation in their own documents.

The texts of statutes and court decisions are, in most countries, free of copyright, and we claim no rights in them. What this licence covers is our compilation: its selection and structure, the version history, identifiers, grades, fingerprints, index and the service itself, which are protected as a database and by copyright.

6. What you may not do

You may not, and may not let anyone else:

  • Extract the corpus in bulk. No systematic downloading, scraping or repeated querying that together reproduces a substantial part of the database, whether in one go or by repeated extraction of small parts.
  • Train or distil a model on it. No use of the service, or of output obtained through it, to train, fine-tune, evaluate for training, or distil any machine-learning model, except the retrieval your own connected model performs while answering your own questions. We reserve text and data mining of the service under Article 4(3) of Directive (EU) 2019/790, by these terms and by machine-readable means (a TDMRep file at /.well-known/tdmrep.json).
  • Resell or redistribute it. No sublicensing, reselling, publishing or making the service or its contents available to people outside your organisation as a service of your own.
  • Get around limits. No circumventing usage allowances, seat counts, jurisdiction entitlements, rate limits or security measures.
  • Misuse it. Anything the acceptable use policy forbids.

7. What you upload, and how we enforce these rules

Documents you upload, drafts and matters are stored for your organisation only; we do not publish them or look at them. We do not screen them in advance. We act on a notice that something is unlawful (see takedown and corrections) and on signs of the misuse listed above, such as use far beyond an allowance or rate limit, which we see from usage counts, not by reading your content.

A person on our staff, not a machine alone, decides every restriction. We start with the least step that stops the problem (a rate limit, then suspending a key or account, then ending the subscription), tell you what we did and why unless the law forbids it, and you can ask us to review it at support@gelex.eu.

8. Not legal advice

The service shows you the law and where it comes from. It does not give legal, tax or other professional advice, and using it does not create a lawyer–client or adviser–client relationship with us. Decisions you take on the basis of the service are yours. If you need advice on your situation, ask a qualified professional. See Not legal advice.

9. AI output and your choice of model

Answers are written by the AI model you connect. You choose that model and its provider, you hold the agreement with them, and you are responsible for that choice: whether the provider may receive what your members ask and upload, where it processes it, and what it does with it. Read the provider's terms before you connect it. We send the provider only what a question needs: the question, your context and the sources the model asks to read.

A model can be wrong, incomplete or out of date even when it is given correct sources. We check the citations an answer makes against the corpus and mark each as verified or not, and check quotations against the source text; we do not check the reasoning. Always read the cited provision before you rely on an answer. How this works is on the AI transparency page.

10. Accuracy of the corpus

We take each text from its official publisher and record where and when we took it. Every document carries a grade (A to D) describing how exactly its structure was captured, and the dates its versions were in force. We work to keep the corpus complete and current, but publishers make errors, publish late, or change texts, and our processing can be imperfect. Where it matters, check the official publication; we link to it. Report an error and we will correct it: see takedown and corrections.

11. Plans, fees and payment

Plans, prices and what each includes are in your plan settings; institutions may have a quoted price instead. Prices to businesses exclude VAT. Fees are paid in advance for the subscription period, by card, SEPA direct debit, MobilePay or bank transfer through our payment provider, Mollie; we never see full card details. We invoice by email.

An allowance not used in a period does not carry over unless your plan says so. A new organisation starts with a free trial; when it ends without a paid plan, it moves to the free plan with its smaller allowance. Nothing is charged without your having chosen a paid plan.

12. Term, renewal and ending

A paid subscription runs for the period you chose and renews for the same period unless cancelled before the renewal date. You can cancel in the app at any time, effective at the end of the current period.

Either side may end the contract if the other materially breaches it and does not put it right within 30 days of written notice; if you end it for our breach, we refund the unused part of the period. We may suspend at once for bulk extraction, model training or a security threat.

While your organisation is open it can export its data at any time. When it closes, we delete its content 30 days later, as described in the privacy policy, including from backups as they expire. On request we confirm the deletion in writing. The audit log is kept apart, as the privacy policy says, because the law or your own audit needs it.

13. Your data, and no training on it

Your questions, documents, drafts and the answers you receive are yours. An answer to you may resemble an answer someone else gets to the same question; that does not make theirs yours or yours theirs.

We do not train, fine-tune or evaluate any AI model on your content, and we run no model of our own. We do not let our sub-processors do so either. We process personal data in your content as your processor under the data processing agreement, which forms part of these terms. Our servers and sub-processors are in the EU or Switzerland, and the service sets no tracking cookies.

14. Availability and changes

We aim for the service to be available at all times and announce planned maintenance in advance in the app. We do not offer a service-level credit unless your contract says so.

We improve the service continuously. If we remove a feature your plan relies on, or change these terms to your disadvantage, we tell you at least 30 days ahead and you may end the subscription with a pro-rata refund before the change takes effect.

15. Liability

  • Each side's total liability under the contract in any 12 months is limited to the fees you paid us for the service in the 12 months before the event that caused the claim.
  • For a breach of confidentiality or of the data processing agreement, that limit is twice those fees.
  • Neither side is liable for indirect loss, such as lost profit, lost business, lost goodwill or loss of data you could have restored from your own copies.
  • We are not liable for the output of the AI model you connect, or for a decision taken on the basis of the service without reading the cited source, beyond what the law makes us liable for regardless.
  • None of this limits liability for intent or gross negligence, for your payment obligations, for a breach of the licence limits in What you may not do, or anything else the law does not allow to be limited, nor a consumer's statutory rights.

16. Law and disputes

These terms are governed by Danish law, without its conflict-of-laws rules and without the UN Convention on the International Sale of Goods. Disputes go to the Danish courts, with the Court of Aarhus (Retten i Aarhus) as first instance. A consumer keeps the protection of the mandatory law of the country they live in and may bring proceedings there.

17. Contact

Questions about these terms: legal@gelex.eu. Support: support@gelex.eu. We answer in English, Danish and German.

Changelog

  1. , version 1: First version.