About SafeLegalAI
SafeLegalAI is the record of AI in legal practice, published by Cognesio LLP: what went wrong, what the official rules say, and what legal tech tools publish about their governance. We maintain a growing set of datasets — the Incident Tracker, the regulatory map and its versioned document registry, and legal tech tools — all free to reuse under CC BY 4.0.
Ownership, plainly
SafeLegalAI is published by Cognesio LLP. LegalAI Space is an affiliated product (a Cognesio LLP product) for law-firm AI governance. For readers, that means LegalAI Space is listed in the tools directory with only its published facts, never scored or ranked, excluded from directory counts and comparisons, and accompanied by a disclosure block whenever it is mentioned.
How this is made
AI does the reading
Language models monitor courts, regulators and vendors across 130 countries, extract what changed, and draft every record and article. That is how a small team keeps a global record current.
An editor signs off
Cognesio LLP's editorial team opens every linked judgment, rule and vendor page, checks each date, quote and status, and owns what publishes. Nothing goes live on a model's word.
The documents are the record
Every cell, row and claim links the official text. Our status labels are an editorial reading of it; when they disagree, the document wins. Corrections are public and dated.
When a page says last checked, it means a human re-opened the linked official documents for that record on that date. Country records marked provisional are AI-researched leads until the Cognesio LLP editorial team has checked every source.
Our summaries point readers to judgments, regulator notices, statutes, guidance and vendor pages; they are not the law and they are not legal advice. Read the full editorial standards and disclaimer.
Open data, open pipelines
Every table we maintain is published as data under CC BY 4.0 and mirrored on Hugging Face, where each row links back to its record here. The pipelines that collect the newer datasets — US court AI orders, the EU AI Act as structured data, the vendor documentation audit — are open source on GitHub, together with their row schemas, the collection rules every crawler follows and the data itself. Anyone can re-run them, check a row against its source, or open an issue. Code is Apache-2.0; data is CC BY 4.0; both disclaim warranty, and the linked official documents remain the record.
Hugging FaceEditorial team
SafeLegalAI is edited by the Cognesio LLP editorial team. The team also builds LegalAI Space, an affiliated product (a Cognesio LLP product). The team verifies AI-assisted research against primary sources and signs off every record and article; accountability sits with Cognesio LLP, not an individual.Editorial team profile →
Corrections & contact
We correct confirmed errors with a dated note on the relevant page. Use the report page to tell us what is wrong, include the page URL and link the primary source you want us to check. You can also email corrections@safelegalai.com.
Seen an AI incident in a courtroom or regulator docket? Send the public record through the report page or to tips@safelegalai.com. We verify against primary sources before anything enters the tracker.
General messages: hello@safelegalai.com. Legal notices: legal@safelegalai.com.