Licensing vs. selling your data

Licensing grants specific, bounded rights to use your material while you keep ownership; selling transfers ownership outright.

Why it matters to a buyer

Buyers get what they actually need — training rights — without owners having to give up their asset.

Why it matters to a data owner

You license, you don’t sell. The material stays yours, you set the terms, and you can license it again.

Current legal status

Standard contract principles; the distinction is in the grant language.

The difference lives in the grant language

A sale is an assignment: the document transfers all right, title, and interest in the material, and after signing, the buyer is the owner and you are not. A licence is permission: the document grants defined rights — to use the material to train models, say — for a defined scope, while title stays with you. The label on the front page does not decide which one you signed; the operative verb does. "Assigns" moves ownership. "Grants a licence" does not.

The trap for owners is sale-shaped language inside licence-shaped documents. A grant that is perpetual, irrevocable, exclusive, worldwide, sublicensable, and "for any purpose" leaves you holding title to material you can never use, license, or price again — economically closer to a sale, without a sale’s price. Common practice treats each of those words as a separately negotiated, separately priced upgrade, not a default.

A worked example: the same dataset, two ways

Imagine an owner with several hundred hours of road-scene footage. Sold outright: one payment, ownership transfers, the buyer can relicense it to others, and the owner has no say in — and no revenue from — anything that happens next.

Licensed instead: the owner grants Lab A a non-exclusive licence to train commercial models, bounded by term. The material stays the owner’s, so next quarter the same corpus can be licensed to Lab B, and an evaluation-only licence can go to a third buyer at a different price. Each licence is a separate grant with its own scope and its own fee. Same footage; one path ends the asset, the other turns it into a catalogue.

When a buyer wants exclusivity

Some buyers will pay for exclusivity — the promise that competitors cannot train on the same material. That is a legitimate ask and a real price lever, but exclusivity is not a sale: common practice bounds it by field of use and term, so an owner can grant, for example, exclusive commercial-training rights for a period while keeping ownership and the right to license again when the term ends. An owner who understands that distinction can charge for what exclusivity is actually worth instead of surrendering title to get the deal done.

Before you sign, check the grant for

Common questions to put to any data deal — practice, not legal advice:

• The operative verb: does the document assign ownership, or grant a licence?
• Exclusivity — and if exclusive, in what field of use and for how long.
• Scope: internal R&D, commercial model training, or "any purpose".
• Sublicensing and transfer: who else could end up holding rights to your material.
• Term and territory, and what happens when the term ends.
• Whether you can license the same material to another buyer tomorrow.
• The payment shape: one-time buyout versus a fee per licence.

What fiund does about it

fiund is a licensing marketplace by design — owners keep ownership, appoint fiund to handle AI licensing while content is active, can withdraw from future licensing, and get paid per licence.

← All rights & provenance guides