For companies
Your company already did the work.
Years of mail, documents, tickets, schemas and ledgers sitting in systems you pay to keep. AI labs pay for exactly that, and most companies never find out because the legal and engineering work looks impossible from the inside.
What actually leaves your building
Business-system records: internal mail and chat between colleagues, documents your team wrote and revised, project trackers and ticket queues, internal wikis, database schemas and structured records, accounting, code repositories.
Not your customer records. Not payment or credential data. Not private messages or human-resources channels. Not anything covered by a third party's confidentiality agreement. Not anything about minors. You write your own additions to that list and they go into the agreement before a single file moves.
Who holds it, and under what
We do, under a licence from you. That is the honest answer and it is different from what an introducer would tell you. It means we carry the obligations: controlled storage, a named de-identification standard applied before anything reaches a lab, a destruction schedule for the raw copies, and the warranties in the downstream licence.
It also means you should read the trust page before you sign anything, including with us.
What you are exposed to
You warrant that you own or hold the rights to what you give us, and that giving it to us does not breach an agreement you are already in. That is a real obligation and it is why the scoping conversation comes before the commercial one: most of the work is deciding what is safely yours.
Where your data is covered by employment or privacy law, there are steps you have to take, such as notifying staff. We tell you what they are, we give you the templates, and we will not let you skip them to close faster.
What it pays
You hold a named share of every licence your corpus earns, fixed in writing before any export. It applies to renewals and to further fields of use, not only to the first deal. You get a statement each quarter showing what was licensed, what was received, what was deducted and how the allocation was calculated, and you have an audit right if you want to check it.
Published programmes in this market pay roughly $50,000 to $2,000,000 per company depending on size, depth of history and field of use. Where your corpus lands inside that range depends on what the manifest actually contains, and we will give you an honest read before you commit any effort.
Whether you qualify
- Roughly 30 to 500 people, including regular contractors.
- Three or more years of continuous operating history.
- Real activity across several systems, not one tool.
- Someone who can say with authority what the company owns.
If you are outside that, tell us anyway. The boundaries are guidance, not a gate, and an unusual corpus is sometimes worth more than a typical one.
What the first hour looks like
- A call to work out what is safely yours and what has to be excluded.
- We size it from your admin consoles, with you watching, so nobody is guessing.
- You get a written read on what it is worth and what the work involves.
- If it is not worth it, we say so. That happens and it costs you an hour.