The major AI companies are paying substantial sums for real-world data to train their models. They license it rather than buy it — you keep ownership, and non-exclusive means the same material can earn again and again.
Your material stays in the processor's secure cloud and the licensee trains on it where it sits — no copy changes hands. Personal and clinical detail is de-identified to a documented, independently reviewed standard first.
Hold the material and choose what goes in. Ownership never transfers.
Establishes the rights position, de-identifies where needed, catalogues and hosts it in secure cloud storage.
Trains on it in place. Non-exclusive, so the same material can be licensed again.
A licence for machine-learning training. Ownership never transfers, and every other right you hold stays with you.
Non-exclusive means the same material licenses to several parties, with payment on each. Recurring income, not a one-time sale.
Training data is consumed by a model — never aired, streamed or resold. Far narrower rights than you are used to negotiating.
We are paid out of the transaction. No charge to qualify, no commitment. If an introducer brought you, any fee is between you and them.
For services firms, agencies, SaaS and professional services — above all for holding companies and portfolio groups.
Your confidential information is de-identified before any licensee sees it. People are scrubbed as a hard floor, by machine and by human review. Client names are replaced consistently, so the shape of the work survives but individual clients are not identifiable. Your own company identity is blinded. Every licensee signs NDA and non-circumvention before access, and each package ships with a manifest of what was done and independent QA against the original.
This is a documented, standardised process with independent review — described to you exactly as it is, not as a guarantee.
The major AI companies are licensing ordinary workflow data: correspondence, customer and support history, project records, finance and operations, production code. There are large live requirements now — this is not a category where you wait for demand. What leaves your systems is de-identified first, and you decide what goes in at all.
Almost nobody holding it has counted it as an asset. The strongest candidates are small and unglamorous, running a decade of history on systems they think of as overhead.
Nobody connects to your systems. No integrations, no installs, no ongoing access — you run a one-time export yourself. That is the entire technical footprint.
These are finite orders. Each closes once its stated volume is met, and licensees fill from the deepest corpora first. State real numbers early and you are in the first tranche.
A group is the highest-leverage version. A holding company licenses each operating entity as a separate event — one decision, many transactions.
Custom systems count too — if a team works in it, it probably qualifies.
For hospital and clinic groups, telehealth platforms, EHR vendors, scribe services and CROs.
Licensees want real clinical material with documented consent covering training use. Demand is worldwide, and records from outside North America and Europe are actively in review. It does not need to arrive de-identified — the processor handles that where you cannot.
Whether records can lawfully leave their country is established before anything moves. We will tell you plainly if the answer is no.
Consent must be documented, not asserted. The processor can de-identify records; it cannot create permission that was never captured. Expect to produce the paperwork — consent form, IRB approval, data-use agreement, or your jurisdiction's equivalent.
The standard is local. Expert determination under HIPAA for US records; elsewhere the local regime — DPDP in India, PDPL in Saudi Arabia, POPIA in South Africa, PIPA in South Korea. Most of what we see is not US-origin, so HIPAA is often not the test.
Longitudinal case packs. One de-identified patient over multiple time periods — notes, imaging, diagnosis, outcomes. Imaging is the strongest addition; complex cases over longer periods are the most wanted.
Records at scale. Single-visit records in volume from an EHR or HIS. Breadth over depth; non-English qualifies.
Visit transcripts. Consultations, telehealth or in-office, any language. Worth materially more with the clinician's note and record entry attached.
Consultation video. In-office or telehealth, recorded anywhere under Western clinical practice.
An archive that earns more than once.
For producers, filmmakers, broadcasters, event companies and studios with footage that has finished earning its keep the first way.
Model quality is bound by the footage available, and there is not enough in the public domain. Processors can only offer what they already hold — material not in the library when an order lands is not considered.
The tail is what most holders do not expect. Because the licence is non-exclusive and your material stays available, the same footage can be selected for separate orders from different licensees — paid on each.
You need the rights to license for this purpose. Old contracts often grant broadcast or distribution while saying nothing about training use, and silence is not permission. Unsure is a normal starting point.
Raw, unaired and B-roll is often the most valuable part of an archive — unpublished, rights in one pair of hands. Independent creators regularly hold stronger material than large production houses do.
If you know archive holders, clinic groups, telehealth platforms or operating companies with years of internal history, bring them to us. We share commission on transactions you originate, under a written agreement covering origination, split and non-circumvention.
No cost, no exclusivity, no volume commitment. Register the introduction in writing, we confirm it, and your position is protected. Get in touch for terms.
A short call to establish what you hold and where your rights position stands. No cost, no commitment.