Data licensing vs selling ownership

A licence is permission with a scope

A licence can limit the content, purpose, duration, territory and users. It can permit one internal evaluation or broader training and commercial deployment. Ownership of the underlying material generally stays with the licensor unless the parties separately agree to transfer it.

Owning source material is not the same as owning every right needed for a proposed use. A company cannot grant permissions it does not hold. Separate third-party components, contractual duties and personal-data obligations from the rights the company can actually license.

A transfer can change future options

An assignment transfers specified IP rights rather than merely allowing a use. The seller may need retained rights to continue operating its product or serving customers. The exact scope should be reviewed rather than assuming everything associated with a dataset is transferred.

Data can involve copyright, database rights, contractual rights and confidentiality. A label such as “full ownership” can hide important distinctions. Before signing, identify what is assigned, what is excluded and what rights the business retains.

Non-exclusive does not mean unrestricted

A non-exclusive licence can allow the owner to license the same permitted material elsewhere while still limiting what each buyer may do. An exclusive licence can restrict future licensing within an agreed scope. Neither term alone explains redistribution, training or deployment rights.

Exclusivity may have opportunity costs and operational consequences. Define the field of use, time period and any exceptions carefully. Do not accept broad exclusivity simply because a proposed payment sounds attractive.

• Can your business keep using and updating the material?
• Can other buyers receive it, and for which purposes?
• Are affiliates, contractors or downstream users included?

AI use needs specific post-training terms

Training can create model weights and other derived artefacts. Agree whether those may be retained, distributed or commercialised. Termination of access to a dataset does not necessarily eliminate knowledge learned by a model.

Discuss deletion of source copies, permitted retention, derived datasets, audit arrangements and handling of breaches. Do not promise that trained models can be fully reversed on request. Get specialist advice on enforceable obligations.

Use a term sheet before debating a headline price

Record the asset, rights holder, use, duration, exclusivity, delivery and payment structure in a short proposed term sheet. This makes comparisons more meaningful and reveals restrictions early.

Commercial terms should account for rights review, preparation, updates and any agreed brokerage fee. AssetGate does not invent prices or act as the purchaser. Any fee and any eventual licence require written agreement.

• List excluded customer, third-party and sensitive content.
• Define evaluation permission separately from commercial use.
• Record what happens to source files and derivatives when the licence ends.
• Have counsel review the rights and privacy position.