National legislation datasets in Europe: sources, licences and formats by country

Every EU member state and the United Kingdom publishes its legislation online for free, and most now offer an API or a bulk XML download: Légifrance and the DILA open-data dumps in France (Licence Ouverte 2.0), legislation.gov.uk in the UK (Open Government Licence v3.0), Gesetze im Internet in Germany (federal statutes are outside copyright under § 5 UrhG), the BOE in Spain and the RIS in Austria (CC BY 4.0). The legal basis for reuse is open almost everywhere, but it is not the same: some sources need a fixed attribution line, some say their consolidated texts are “for information only”, and some apply national privacy rules on top. Arieta Data’s 2026-11 snapshot holds 350k+ national legislation documents from 15 countries, with the source, licence and attribution on every record.

What is a national legislation dataset?

A national legislation dataset is a machine-readable collection of a country’s statutes and secondary legislation (acts, decrees, regulations, statutory instruments, ministerial orders), with their metadata and ideally their text split into articles or sections. Most national portals publish consolidated texts, which merge an act with its later amendments. In most countries those consolidations are for information only, and the authentic version is the one in the official gazette. Some countries, such as Denmark (with its lovbekendtgørelser) and Poland (with its teksty jednolite), publish official consolidations too.

For AI work, four properties matter more than the raw text:

1. A stable id per act, ideally an ELI, so that amendments, citations and EU transposition records join without string matching. See CELEX, ECLI and ELI explained.
2. Structure: article, section and paragraph boundaries that survive parsing. See How to chunk legal documents for RAG.
3. Dates: adoption, entry into force, publication, and when the text was last updated.
4. Terms: the licence, the attribution text and any condition that must travel with the data.

Which official source does each country use, and under what terms?

The table lists the official sources behind Arieta Data’s 15 national lanes, with the terms each publisher states. Document counts are from snapshot 2026-11; article counts for France and the UK are estimates from a 2 percent sample.

Country	Official source	Terms of reuse	Documents (2026-11)	Articles
France	Légifrance / DILA	Licence Ouverte 2.0: name the source and the date of last update	117,321	approx. 719k
United Kingdom	legislation.gov.uk (The National Archives)	Open Government Licence v3.0; personal data is outside the licence	79,497	approx. 776k
Lithuania	Teisės aktų registras (TAR), via data.gov.lt	CC BY 4.0 (data.gov.lt dataset 2613, checked 14 Nov 2026); legal acts outside copyright	29,325	21,722
Croatia	Narodne novine	Official texts outside copyright	27,077	475,540
Netherlands	Basiswettenbestand (wetten.overheid.nl)	No copyright in statutes (Auteurswet art. 11) and no database right for them (Databankenwet art. 8)	22,310	415,754
Poland	Sejm ELI API / ISAP	Official acts excluded from copyright (art. 4 Prawo autorskie)	18,534	212,793
Denmark	Retsinformation (Civilstyrelsen)	Free, non-exclusive, unrestricted right of use, commercial use included; no implied endorsement (Retsinformation API terms, checked 14 Nov 2026); statutes outside copyright (§ 9 ophavsretsloven)	10,715	171,023
Austria	RIS (Bundeskanzleramt)	CC BY 4.0	10,212	147,805
Slovenia	PISRS	CC BY 4.0	9,362	209,151
Czech Republic	e-Sbírka	No copyright work, no protected database, no sui generis database right (national data catalogue terms of use, checked 14 Nov 2026)	8,700	97,078
Spain	BOE	AEBOE reuse conditions: fixed attribution, “texto consolidado de carácter meramente informativo”	6,833	205,167
Germany	Gesetze im Internet (BMJV / BfJ)	Official works outside copyright (§ 5(1) UrhG); no licence stated	5,921	107,552
Sweden	Riksdagens öppna data	Free use; always name Sveriges riksdag as the source	5,320	108,864
Finland	Finlex	CC BY 4.0	5,265	93,600
Italy	Normattiva	CC BY 4.0 (in force since 1 January 2026; the earlier experimental phase was CC BY 4.0 NC until 31 December 2025, and Arieta Data’s records were all retrieved in 2026)	1,553	19,418

Two things stand out. First, the size of a lane says little about the size of a country’s law: Lithuania’s lane is mostly short ministerial orders (įsakymai) and resolutions, which is why it has more documents than articles, while Spain’s BOE lane is long consolidated acts, averaging about 30 articles each. Second, the open licences are not interchangeable. CC BY 4.0, the Licence Ouverte and the OGL all permit commercial reuse with attribution, but each wants its own attribution form, and none of them makes a consolidated text authentic.

How do the big three compare: France, the UK and Germany?

France, the UK and Germany are the jurisdictions buyers ask about most, and each has a different model.

France. The DILA publishes the LEGI database of consolidated codes, laws and regulations as open data, and the Légifrance API is available through the government’s PISTE gateway. Both are under the Licence Ouverte 2.0, which permits commercial reuse worldwide provided you name the source and the date of last update, and do not suggest an official character. In Arieta Data the French lane is consolidated arrêtés and décrets plus laws and codes: 117,321 documents, the largest national lane. French case law is a different matter. Article L111-13 of the Code de l’organisation judiciaire bans reusing the identity data of judges and court clerks to evaluate, analyse, compare or predict their professional practice, and that ban carries criminal penalties.

United Kingdom. legislation.gov.uk publishes UK, Scottish, Welsh and Northern Ireland legislation, original and revised, with XML for every item, under the Open Government Licence v3.0. The OGL lets you exploit the information commercially; it excludes personal data and any suggestion of official endorsement. Arieta Data’s UK lane has 79,497 documents, of which 56,428 are UK Statutory Instruments and 3,519 are Public General Acts. UK case law is a separate question, and not a happy one for AI teams: the Open Justice Licence does not cover computational analysis, including machine learning, and The National Archives requires a separate licence for it. Arieta Data does not offer UK case law; see the UK case-law lane.

Germany. Gesetze im Internet publishes consolidated federal law, with a table of contents and per-act XML that its notes describe as suited to automated processing. It states no licence, because statutes and official decisions are “amtliche Werke” without copyright under § 5 UrhG. The consolidations are published as non-official texts. Arieta Data’s German lane holds 5,921 federal acts and ordinances. German Land law is published by each state, separately.

What conditions travel with the data?

Open terms still carry obligations, and they bind you, your dataset documentation and, where you redistribute, your customers.

Condition	Where it applies	What to do in a pipeline
Attribution in a set form	Spain (“Basado en datos de la Agencia Estatal Boletín Oficial del Estado” for derived works, with a link to boe.es), Sweden (name Sveriges riksdag as the source), every CC BY source	Keep a per-record attribution field and carry it into your datasheet and model card
Date of last update	France (Licence Ouverte), Spain (BOE)	Store the source’s own update date, not only your fetch date
Not the authentic text	Almost every consolidation	Label records as non-authentic; never present output as the official text
No endorsement	UK (OGL), Denmark, Sweden, Spain	Do not use the publisher’s name or logo to imply backing
Privacy	Finland (Finlex forbids combining data in a way that violates privacy), the OGL’s personal-data exclusion	Legislation holds little personal data, but keep these terms in your licence flow-down
CC BY stays CC BY	Austria, Finland, Slovenia, Lithuania, Italy (Normattiva)	Verbatim CC BY text can be redistributed by anyone who receives it, whatever a reseller’s terms say

The last row matters if you license data from a reseller, Arieta Data included. CC BY 4.0 forbids adding terms that restrict the licensed rights, so a reseller’s “no redistribution” clause can bind only its own contribution (structure, metadata, history, the citation graph), not the CC BY text itself. Arieta Data marks CC BY records in every shard.

Copyright is rarely the obstacle for legislation, because statutes are excluded from copyright in most European countries. Whether text and data mining opt-outs and the AI Act copyright policy still matter is covered in EU copyright, TDM opt-outs and the AI Act for legal data.

Why not scrape each portal yourself?

You can, and the licences generally allow it. The work is in the details that differ per country:

• Access channels. Some publishers offer an API (the Sejm ELI API, the RIS OGD API, Retsinformation’s API), some offer bulk XML (DILA, Gesetze im Internet), and some publish rules on pacing or ask to be told about bulk downloads. Several block generic crawlers in robots.txt while inviting use of their documented XML or API channel.
• Formats. Each country has its own XML schema, numbering of articles and paragraphs, and way of marking amendments.
• Identifiers. Some countries publish ELIs and some do not; national ids are all different.
• Change. Consolidations are rewritten whenever an amendment applies, so a one-off crawl goes stale within weeks. Tracking change needs version history and a way to handle withdrawals.

Multiply this by 15 countries and the engineering cost dwarfs the licence fees, which are zero. That trade-off is set out in EU legislation datasets for LLM training, and for web-scale crawls in Common Crawl vs curated legal data.

What about national case law?

National case law is published by most of the same countries, often under the same open terms, but it names or pseudonymises people, and some sources add their own restrictions. The French judge-analytics ban and the UK Find Case Law licence are the best-known examples. Arieta Data includes national case law, delivered with personal data masked. CJEU case law is included in the EU lane, and is not in the free open sample; a CJEU case-law excerpt is available under an evaluation licence on request. Italian Corte costituzionale decisions are CC BY-SA 3.0 (share-alike); they are delivered masked as a separate CC BY-SA file (details).

Key facts

• Arieta Data snapshot 2026-11 holds 357,945 national legislation documents from 15 countries, alongside 61,410 EU documents: 419,355 documents in all.
• The largest national lanes are France (117,321 documents), the UK (79,497), Lithuania (29,325) and Croatia (27,077).
• The open terms are Licence Ouverte 2.0 (France), OGL v3.0 (UK) and CC BY 4.0 (Austria, Finland, Slovenia, Lithuania, Italy); Germany, the Netherlands, Poland, Croatia and Denmark rely on statutory exclusions from copyright.
• Almost every national consolidation is non-authentic: only the official gazette is.
• UK case law is not covered by the Open Justice Licence for AI training; French case law carries a criminal ban on profiling judges and clerks.

Frequently asked questions

Is French legislation free to use for AI training?

Yes, in general. The DILA’s LEGI dataset and Légifrance content are published under Licence Ouverte 2.0, which permits commercial reuse worldwide if you name the source and the date of last update and do not suggest an official character. The stricter rules apply to French court decisions, not to legislation.

Can I train a model on legislation.gov.uk?

Yes. UK legislation on legislation.gov.uk is published under the Open Government Licence v3.0, which permits commercial and non-commercial use with the attribution “Contains public sector information licensed under the Open Government Licence v3.0”. UK court judgments on Find Case Law are different: computational analysis, including AI training, needs a separate licence from The National Archives.

Does German federal law have a licence?

No licence is stated, because none is needed for the texts: statutes, ordinances and official decisions are outside copyright under § 5 UrhG. Gesetze im Internet offers XML for automated processing and labels its consolidations as non-official.

Which countries publish ELI identifiers?

The EU and a growing number of member states. The EUR-Lex ELI register lists, among others, Belgium, the Czech Republic, Denmark, Germany, Spain, France, Croatia, Italy, Lithuania, Luxembourg, Austria, Portugal and Finland as implementing it, with others partial or in progress; legislation.gov.uk uses ELI-style URIs. Where a source supplies an ELI, Arieta Data keeps it on the record next to the national id.

Where Arieta Data fits

Arieta Data is a licensed European legal corpus from Arieta Labs Ltd, the company behind LexAPI. Snapshot 2026-11 covers EU legislation, CJEU case law and national legislation from 15 countries, in 24 languages (EU documents are almost all in English; the national lanes cover 14 national languages): 419,355 documents, approx. 4.07M articles and approx. 2.05B tokens, split at article level with source, licence and attribution on every record. It ships as the corpus, the version history and the citation graph, with per-lane detail on the coverage page and terms on the licensing page.

Request a sample for the countries you need. We reply within one business day, and a custom cut by jurisdiction, language, date range or document type is ready within 5⁠–⁠6 business days.

Not legal advice. Publishers change their terms; check each source’s current wording and take your own legal advice before training on any dataset.