Mobile Proxy Legal Data
Case law and regulation are published deliberately, jurisdiction by jurisdiction, through services that were built for lawyers rather than for crawlers. The work is coverage and citation integrity, not access.
- Every jurisdiction has its own service — BAILII, Legifrance, EUR-Lex and CourtListener are four different worlds, not four mirrors.
- Judgments carry personal data — several jurisdictions pseudonymise deliberately, and re-identifying is a serious matter.
- Citation is the hard problem — the same judgment carries different identifiers in different systems.
- These are public services — patient, light-touch collection is both correct and more reliable.
Reach each jurisdiction’s own legal database.
Record the identifier each system uses.
Legal Data Collection
Legal material is published deliberately, jurisdiction by jurisdiction, through services built for practitioners rather than for crawlers. Access is rarely the problem. Coverage and citation integrity are.
The services share almost nothing structurally. England and Wales has BAILII and the National Archives judgment service; France has Legifrance; the European Union has EUR-Lex; the United States federal courts sit behind PACER, with CourtListener and the RECAP archive providing free access to a substantial subset. Each has its own identifiers, its own coverage depth, its own idea of what constitutes a document, and its own search semantics. There is nothing to build once and point at five countries.
| Jurisdiction | Primary services | Access character |
|---|---|---|
| England and Wales | BAILII, National Archives judgments | Open, free, rate-sensitive |
| France | Legifrance | Open, with pseudonymised judgments by design |
| European Union | EUR-Lex | Open and well structured, with documented interfaces |
| United States (federal) | PACER, CourtListener, RECAP | Charged per page retrieved; free mirrors cover a large subset |
Note the shape of the constraint in the American case. PACER is not blocking you; it is charging you. An untargeted crawl there is expensive rather than refused, which makes a narrower query the correct response — and a wider address pool an actively wrong one.
Robots.txt and request pacing differ by source
BAILII and the National Archives judgment service both publish a robots.txt that excludes specific search and admin paths while leaving the judgment text itself open; ignore it and expect the same throttling any other public service applies to a client that ignores its stated rules. EUR-Lex exposes a documented webservice with its own SOAP and REST interfaces, which is the correct entry point rather than the rendered HTML — scraping the page when a service endpoint exists just recreates work the publisher already solved. CourtListener's API sits behind an API key with its own per-key rate limit, layered on top of whatever the IP itself is doing, so a shared or previously-flagged address can leave a well-behaved client throttled for reasons that have nothing to do with its own request rate.
Case Law Research
The hard problem in case law is not retrieval, it is identity. The same judgment appears under a neutral citation, one or more law-report citations, a court docket number and a database-internal identifier, and different systems prefer different ones. Cross-jurisdiction research lives or dies on resolving those to a single entity, and in practice that work dwarfs the collection itself.
neutral_citation … report_citations [ … ] # frequently several, from different reports docket_number … source_id … # the database's own identifier court … decided_on … source bailii | legifrance | eurlex | courtlistener fetched_at 2026-08-24T09:14:02Z
The second discipline is versioning. Judgments are corrected and re-issued, and legislation is amended continuously — consolidated texts change while the address stays the same. Recording which version you read, and when, is what makes it possible to answer what the law was on a given date rather than only what it is now.
The citation formats themselves carry no shared logic
A neutral citation such as [2024] EWCA Civ 12 tells you the court and the year but nothing about where the text lives; an ECLI identifier like ECLI:EU:C:2023:456 encodes country, court and year in a fixed structure the EU adopted precisely so member states' citations could be compared programmatically; a US federal docket number identifies the case but not a specific filing within it, which is what RECAP's document numbering is for. Building a lookup table that maps each format to the fields it actually guarantees, rather than treating all four as interchangeable strings, is what makes automated cross-referencing possible at all.
Regulatory Information Gathering
Regulation is the higher-volume half of this vertical and the one with more commercial demand. National gazettes, regulator publications, consultation papers, guidance notes and enforcement decisions all move faster than case law and matter more directly to businesses trying to stay compliant across several markets.
Here a local exit is more often genuinely necessary. Some national services restrict or degrade access from foreign address ranges, and several render interfaces, search behaviour and document availability differently depending on where the visitor appears to be. Reaching a jurisdiction’s service from inside that jurisdiction is simply the most reliable way to see what practitioners there see, and it removes a class of intermittent failure that is otherwise very hard to attribute.
- Watch the index, not the documents — Gazettes and regulators publish to a schedule. Poll the listing and fetch only what is new.
- Capture the consultation as well as the outcome — The draft and the responses frequently signal the direction months before the final instrument.
- Keep the original language — Store the source text and translate downstream. A translated legal text is a summary, not evidence.
Enforcement decisions are not case law
A regulator's enforcement decision, guidance note or consultation response is persuasive rather than binding in the way a court judgment is, and it can be revised or superseded without any of the procedural markers a legal system attaches to overruling a precedent. Tag enforcement material separately from judgments in the dataset from the start — merging the two categories because both happen to be legal text loses the distinction a lawyer using the dataset will need immediately, and it is far cheaper to keep them apart during collection than to sort them apart afterward.
Cross-Jurisdiction Data
Comparative work is the reason to build this at all, and it introduces a duty that pure collection does not. Several jurisdictions publish judgments in pseudonymised form specifically so that individuals are not permanently searchable — French practice is a well-known example. Collecting the published text as published is ordinarily fine. Attempting to re-identify parties, or aggregating across sources in a way that defeats a pseudonymisation one source applied deliberately, is a serious matter.
That is a question for your own legal advice rather than an engineering decision, and it is worth raising before the pipeline is built. A cross-jurisdiction dataset can quietly recreate the searchability that one jurisdiction spent effort removing, without anyone intending it.
Where the dataset is stored is a second question
Collecting a pseudonymised French judgment inside France and then mirroring the full dataset onto infrastructure outside the EU raises an international-transfer question that is separate from the collection question, even though the two get conflated in practice. The safer default for a cross-jurisdiction legal archive is to keep storage aligned with the strictest regime among the sources it draws on, rather than the most permissive one, and to treat that as a data-architecture decision made before collection starts rather than something to retrofit once the archive already exists.
These are public services, frequently run on modest budgets, and they are used by people doing their jobs. Patient, single-threaded, off-peak collection is both the correct behaviour and — because it never triggers a defensive response — the more reliable one.
For collection discipline see web scraping best practices, and for the neighbouring public-sector sources, government data.
Cover Each Jurisdiction From Inside It
Live PXM2 locations — pick the jurisdictions your research covers and reach each service locally:
France
India
Poland
Frequently Asked Questions
Where is case law actually published?
In national and supranational services that share almost nothing structurally. England and Wales has BAILII and the National Archives judgment service, France has Legifrance, the European Union has EUR-Lex, and the United States federal courts sit behind PACER with CourtListener and the RECAP archive providing free access to a large subset. Each has its own identifiers, its own coverage depth and its own idea of what counts as a document.
Is there a cost or access barrier?
Sometimes, and it is worth knowing which kind. Some services are free and open. Some are free to read but rate-limited. PACER charges per page retrieved, which makes an untargeted crawl expensive rather than blocked. Where a fee applies, the correct response is a narrower query, not a wider address pool.
What about personal data in judgments?
It is real and it is regulated. Several jurisdictions publish judgments in pseudonymised form precisely so that individuals are not permanently searchable, and French practice is a well-known example. Collecting the published text is ordinarily fine; attempting to re-identify parties, or aggregating across sources in a way that defeats the pseudonymisation, is a serious matter and one for your own legal advice rather than an engineering decision.
What is the hardest part in practice?
Citations, by a distance. The same judgment appears under a neutral citation, a law-report citation, a docket number and a database-internal identifier, and different systems prefer different ones. Cross-jurisdiction research lives or dies on resolving those to a single entity, and that work dwarfs the collection itself.
Why would a local exit matter for legal sources?
Some national services restrict or degrade access from foreign address ranges, and several render interfaces, search behaviour and document availability differently depending on the visitor’s apparent location. Reaching a jurisdiction’s service from inside that jurisdiction is simply the most reliable way to see what practitioners there see.
Related Mobile Proxy Guides
Legal and government sources overlap heavily — both are public services published one jurisdiction at a time.
Business use cases
Core mobile proxy guides
Cover Each Jurisdiction Locally
Dedicated 4G/5G modems with unlimited bandwidth and unlimited rotations — carrier IPs inside the jurisdictions your research covers.
Get a Mobile Proxy