Insights · Data licensing

Selling data or licensing the right to use it?

Business conversations often refer to selling data. That expression alone says little about the actual agreement. What matters is which uses are permitted, which rights remain with your organisation and what obligations the contract places on each party.

1. Define the permitted purpose

A licence specifies the conditions under which another party may use a particular dataset. Possible purposes include research, model development or a limited technical evaluation. These purposes should be described explicitly. A broad phrase such as 'for AI' can leave important questions unanswered.

Discuss the processing activities involved, who will perform them and the outputs that are expected. Training, evaluation and onward distribution of a dataset are different activities. The agreement should make clear which of them, if any, are authorised.

2. Distinguish continued use from data delivery

With appropriately agreed terms, your organisation may retain its original archive and continue using it in its own operations. That does not mean that no copies will be transferred to a licensee. Data can be made available through several technical arrangements.

Address two separate questions: which uses remain available to your organisation, and what data will the other party actually receive? Define the approved dataset, access permissions, delivery method and safeguards. The technical arrangement should reflect the scope agreed in the contract.

3. Make exclusivity a deliberate choice

A non-exclusive licence may leave room for other uses or additional licensing agreements. Whether this is possible, and to what extent, depends on the contract and the underlying rights. Non-exclusive should therefore not be treated as a promise of unrestricted freedom.

For an exclusive agreement, its precise scope deserves particular attention. Does exclusivity apply to a specific application, industry, region or every possible use? Consider the potential effect on customer commitments, partnerships and future business opportunities.

4. Consider duration and outputs together

A contract term does not automatically determine what happens to outputs already created when the agreement ends. For an AI application, the parties should clarify whether a trained model may continue to be used and what deletion, evidence or restriction requirements apply.

Do not assume that information can simply be withdrawn from a trained model. Required measures need a technically meaningful and clearly defined contractual basis. Onward disclosure, sublicensing and the involvement of additional service providers also belong in this discussion.

5. Compare remuneration with the scope granted

A payment can only be assessed in relation to what is being provided. Which datasets are included? Are updates, preparation work, technical support or additional documentation required? What rights does the licensee receive, and how broad are those rights?

Compare offers using consistent criteria. A higher headline payment may come with wider obligations or greater restrictions on future opportunities. Payment timing, acceptance criteria and any intermediary remuneration should also be clarified before the parties commit.

Use a short decision brief to compare options

Before deciding, summarise the approved dataset, proposed recipient, permitted purpose and principal terms in a short brief. List unresolved questions and the people responsible for addressing them. This does not replace the agreement, but it can reveal misunderstandings between management, IT and legal advisers. It is particularly useful when proposals involve different rights or preparation duties, making clear what a commercial comparison covers and what still requires review.

6. Establish the rights before agreeing terms

A carefully worded licence cannot replace rights that are missing. Customer rights, confidentiality obligations, personal information and third-party contributions need to be assessed for the particular dataset. Calling an arrangement a licence does not itself make the proposed use permissible.

BRIDGERS acts as an intermediary for potential data partnerships. The starting point is a general description of the archive, followed by discussion of requirements, rights and commercial terms. Final approval must relate to the actual use being agreed.

What’s in your data archives?

A short description is all you need to start. No raw data required.

Assess your data