A legal AI model built through federated learning
Privacy-preserving, collaborative training across rights-cleared private legal sources—without sharing or centralizing raw data.
Law firms, companies, and individuals retain control of their data while enabling the model to learn from it—and share in the revenue it generates.
- Documents you upload
- None. Not encrypted, not temporarily, not at all.
- What leaves your environment
- Model updates — the pattern, not the paper.
- What you receive
- A share of metered usage, on a published formula.
- Leaving
- Any time. Nothing of yours was ever ours to keep.
The world's legal knowledge is locked away
The most valuable legal knowledge—negotiated clauses, practical judgment, and decades of deal history—sits inside private document systems.
It cannot be published or shared with outside AI models because it is confidential. As a result, legal AI learns from public records but misses the private expertise that drives real-world decisions.
Organizations are sitting on immensely valuable knowledge, yet their archives remain cost centers.
The learning is federated. The data isn't.
Your documents stay where they are
You don't upload anything. Training runs where your files live, under your control. OpenLegal never has your documents.
Only the learning joins the model
What comes out of training is the pattern, not the paper. It's engineered and audited so that no one — including us — can pull your text back out of it.
You get paid when the model works
Every product built on OpenLegal pays to use it. Metered usage, a published formula, a payment every quarter.
The knowledge already exists. It's been in the drawer for a hundred years.
OpenLegal opens the drawer without taking anything out.
