Licensing you can document item by item.
When counsel asks where your training data came from, you should be able to answer for every item. We license data in a way that makes that possible.
Data licensing, with proof.
Questions about training data rights now come from counsel, customers, partners and regulators. Answering them requires more than a general warranty: you need to know, for each item, who created it, what they agreed to, what terms apply and whether it shows people or property that need releases. HUMXN binds AI-training permission, licence terms and any model or property releases to each item at intake, and those records travel with the item into your delivery.
Licences are granted per dataset with terms visible per item. Non-exclusive licences suit broad training data. Exclusive licences are available for commissioned work and for catalogue items whose creators agree, and exclusivity is recorded on the item so it is not offered to another buyer within the licensed scope. Standing orders extend the same terms to new items over time. Creators are paid for licensed work, which is what makes ongoing supply sustainable.
Provenance is cryptographic rather than descriptive. Each item has an Ed25519-signed provenance record, events are written to a hash-chained signed audit trail, and records are anchored with OpenTimestamps in Bitcoin, so their timing can be independently checked. A C2PA manifest is attached where the format allows, and licensed image deliveries carry forensic delivery marks. Your signed receipt lists every item by ID and hash. These records are designed to support your compliance work; they are not legal advice.
Built to your specification.
Every engagement starts from a written spec and a pilot batch. These are the most common requests we source for data licensing.
Per-item consent and terms
AI-training permission and licence terms bound to every item and included in the delivery.
Releases on file
Model and property releases attached to items showing identifiable people or private property.
Exclusive licences
Exclusivity for commissioned work and willing creators, recorded so items are not resold within scope.
Signed provenance
Ed25519-signed records per item, a hash-chained audit trail and OpenTimestamps anchoring.
Signed receipts
A receipt listing every delivered item by ID and hash.
Forensic delivery marks
Marks on licensed image deliveries that tie a copy to the delivery it came from.
01Warranties are not records
A general rights warranty does not tell you who agreed to what. Per-item consent and terms do.
02People in data
Images, video and audio of identifiable people need consent on file. Items without it are not offered for training.
03Proving what you trained on
Hashes, signed receipts and timestamped provenance let you show the exact contents of a training set later.
04Origin disclosure
Each item records whether it is human-created, AI-assisted or AI-generated, so you can describe your data accurately.
Data licensing: asked often.
What rights does a HUMXN licence grant?
Each delivery comes with a licence covering the use agreed in your order, typically training and evaluating models, and the per-item terms set by each creator. Those terms, any releases and any restrictions are listed with the delivery. Your counsel should review the licence against your intended use; we do not provide legal advice.
What is the difference between exclusive and non-exclusive licences?
A non-exclusive licence lets the same item be licensed to other buyers. An exclusive licence means the item is not offered to another buyer within the licensed scope, and that status is recorded on the item. Exclusivity is standard for commissioned evaluation sets and available elsewhere when creators agree. It increases price.
How can we prove what data we trained on?
Your signed receipt lists every delivered item by ID and hash. Each item also has an Ed25519-signed provenance record, events are kept in a hash-chained signed audit trail, and records are anchored with OpenTimestamps in Bitcoin so their timing can be independently checked. Together these let you show the exact contents of a dataset later.
Does this make our model compliant with AI regulations?
No dataset can do that on its own. Per-item consent, origin classification, provenance records and receipts are designed to support the documentation your compliance work may require under the rules your organization must consider. Whether and how they meet a specific obligation is a question for your counsel.
How are creators paid for licensed data?
Creators set terms for their work and are paid when it is licensed. Experts who create or review commissioned data are paid automatically after identity and tax verification. Paying the people who make the data is what allows consented supply to continue through standing orders as your needs change.
Tell us what your model needs to learn.
Send a brief in five minutes. A data lead replies within one business day with questions and a first sourcing plan.