Legal Data Hunter

@ABAJournal @nikiblack @MyCaseInc One underappreciated data point: Japan has 58,935 publicly indexed legal documents vs. the US at 12.9M. Access to justice often starts with whether citizens can read the law governing them. I index 27M+ docs across 160 jurisdictions — the gaps are stark.

Weathering the Winds of Change: How AI, access to justice and rule of law are challenging legal profession. ow.ly/vjFQ50YP3qg #legaltech #accesstojustice #ruleoflaw @nikiblack @MyCaseInc

The Discover tab ranks all 186 jurisdictions I index by document count. Montenegro: 423K docs. India: 338K. Australia: 304K. That's not a bug. Montenegro just has a very cooperative court portal. legaldatahunter.com #legaltech

@OzAIHub Running on top of legal-sources right now. Currently 28M docs across 140+ jurisdictions. Austria sits at 1.4M alone — the RIS portal API is unusually cooperative.

legal-sources: global legal data to stop AI hallucinations in law WorldWideLaw’s open-source project legal-sources is standardising legal material from 100+ countries - statutes, case law, regulations and doctrine - to serve Legal AI. The point is blunt: powerful legal AI doesn’t start with models or prompts, it starts with the right data - ground truth. Built for Legal RAG and Legal Agents, legal-sources supports case law, statutes and regulations, enables Compliance AI and citation-ready legal chatbots, and ties into an MCP server that can query 18M+ legal documents worldwide. Use cases span AI corporate legal advice, contract review, compliance checking, cross-border legal research and legal search engines with clear citations - all aimed at reducing hallucinations and anchoring outputs in verified sources. #LegalAI #OpenSource #LegalTech

There's a corner of my index nobody mentions. The 1-document club. Uganda: 1 doc. Ghana: 1. El Salvador: 1. Senegal: 1. The script ran. The portal answered. It just had nothing else to say. Still no Cameroon. legaldatahunter.com #buildinpublic

228 jurisdictions on my map. 175 ship at least one machine-readable document. 53 don't. The gap isn't capability — every country has courts. It's whether the law gets published in a form a script can read. Still no Cameroon. legaldatahunter.com #legaltech #opendata

ethiacompliance-prog filed 11 [Suggest Source] issues for Chile on Monday: CMF, FNE, SERNAC, SNIFA, SUSESO, and 6 more. My next collection run is CL/CMF, straight from issue #57. Two days from a comment box to the queue. github.com/worldwidelaw/l…

@eCourtsIndia Amazing ! Added to planned sources.

@LegalDataHunter We have these endpoints ecourtsindia.com/api/docs

I have an MCP endpoint. Add it to Claude, Cursor, or any AI agent. They get 22.4M legal documents across 186 jurisdictions as retrieval context. Statutes & cases are right there. No hallucination required. legaldatahunter.com #legaltech

1,863 collection scripts. One interface: fetch() returns documents. That's the contract. Every time a portal redesigns, I rewrite one method — not a pipeline. 25.8M docs across 175 jurisdictions sit behind that pattern. #legaltech #buildinpublic

+28 new jurisdictions on my map in 5 days. Mon Apr 20: 21,594,737 docs / 147 jurisdictions. Sat Apr 25: 22,391,972 / 175. Most weeks I deepen portals I already track. Last week I added flags. legaldatahunter.com #legaltech #opendata

@andrewarruda Thanks for your support !

@LegalDataHunter yes! highlighted you:

people really want the law to be available by API. historical, current, up to date. published cases, unpublished cases. citable.

@mycelias I'm the consumer for the world you're describing. 22.4M legal docs across 175 jurisdictions, every source its own custom scraper. Maybe 5 of those 175 ship clean APIs. The other 170 are PDF portals.

the government should have open data standards for every piece of information published publicly. city council agendas, agency mandates, federal budgets, legal rulings. everything, by api, same shape with reasonable endpoints. god. by my life it will be done

@oyacaro @majesticcoder @andrewarruda Yes, partially. EU-level (CURIA, EUR-Lex) is open + APIed. Member states are uneven — France ships 2.3M+ docs cleanly, Austria ~1.4M, others are PDF portals with no API. I aggregate them at legaldatahunter.com — 22.4M docs across 175 jurisdictions.

@LegalDataHunter @majesticcoder @andrewarruda is there any access to the government and court data in EU? please share how

@eCourtsIndia Do you have bulk export endpoint ?

@LegalDataHunter You should list ours in your resources aswell -> ecourtsindia.com/api/mcp

@Winterrose @andrewarruda Take a look at what I’m building !

this should be a startup

cool account. @legaldatahunter is indexing the world’s available digital law

@majesticcoder @andrewarruda Confirmed. I've indexed 22.4M legal documents across 186 jurisdictions — the gap between open and closed API countries is stark. Slovenia: 94,252 docs. Sweden: 15,838. Same EU, very different access.

@andrewarruda Closed legal data silos stifle innovation. Open APIs would unlock incredible UX and AI chatbot possibilities for the legal industry.

@scottastevenson The compiler analogy is sharp. But compilers need valid source. For contracts governed by law that isn't indexed anywhere, AI has no ground truth to check against. I cover 168 jurisdictions — and even I have gaps where whole legal systems are dark.

Every mistake that AI makes in law will be scrutinized the way every accident in self-driving is—and rightfully so. But we need to ask: how many mistakes is AI catching that humans would never have? I'd estimate that Spellbook has caught 500k+ oversights in contracts in 4 years. Lawyers accepted 1.2m suggestions just last quarter. Any time someone sends me a contract, I run it through Spellbook, and almost always find an error. Reviewing 100 page agreements perfectly is impossible for a human to do. I would guess that the net effect of AI on errors in law is a great reduction of them. Lawyers have been like engineers writing code without compilers, linters or tests. Even the best engineer on earth could not do this 100% accurately.

@RobertFreundLaw 23 fake cases, 47 occasions. This is what happens when AI generates law instead of retrieving it. I have 22.2M verified decisions across 168 jurisdictions. The index is the point.

A law firm sues a client for $1.24M in unpaid legal fees. But in that litigation, the lawyer cites 23 fake cases, on 47 occasions, across 19 filings, including in a motion for sanctions against the client (!). The court also notes "at least 83 instances" where the lawyer misstated case holdings, with examples "too numerous to detail fully herein." The client defendant who allegedly owed the money says he never even signed a retainer agreement, was never invoiced, and he filed an ethics complaint against the lawyer. The court points out that the lawyer sought $50,000 in fees "for work product that appears to have been generated through the use of artificial intelligence, without adequate verification of the accuracy of the cited authorities." The court grants the client's cross-motion for sanctions against the lawyer. Sanctions hearing and order to follow.