Insights
You can license company data and still keep the company.
Learn how to license company data to AI buyers without selling the business. Compare exclusivity, training and retrieval rights, field of use, and term.
This is general information, not legal advice. Have counsel review any data licensing decision.
Licensing is not the same as selling.
Many companies assume there are two options. Keep the data, or sell the business that holds it. There is a third option. License the data and keep the company.
A license gives a buyer permission to use defined material for defined purposes. It usually has a term. It can be limited by field, geography, and delivery method. The company stays the owner.
This article explains how that works. It covers the three paths, exclusivity, training and retrieval rights, field-of-use and term limits, what stays with the company, and a checklist. It also covers when licensing makes sense.
Three paths, compared.
There are three common ways data leaves a company’s control. They are different transactions.
Licensing data. The company grants permissions. It keeps title. The buyer receives access to a defined set of material, under stated limits, for a stated period.
Selling data. The company transfers the dataset or the rights in it. After closing, the buyer may control how it is used. The company may keep only what the agreement reserves.
Selling the company. The buyer acquires the entity, its assets, and its liabilities. Data goes with the company, along with every restriction attached to it.
Each path has different consequences:
- Ownership: licensing keeps it. The other two move it.
- Reversibility: a license can end. A sale generally cannot be undone.
- Ongoing duties: licensing may require delivery, support, and compliance. A sale may end them.
- Diligence: a company sale reviews the whole business. A license reviews the licensed material and your authority to grant rights.
If the goal is to create value from data while the business continues, a license is often the place to start.
Exclusive and non-exclusive licenses.
Exclusivity is a commercial term. It should never be a default.
A non-exclusive license lets the company license the same material to others. It also lets the company keep using it. This is the simpler structure. It leaves room for several buyers.
An exclusive license limits who else may receive the licensed material. A buyer may ask for it because it wants a competitive advantage. That request deserves a clear answer on what it covers.
Exclusivity can be narrowed in several ways:
- By content: only one collection, not everything the company holds.
- By field: only one use, such as a single product category.
- By time: only for a fixed period.
- By buyer type: only against named competitors or a defined class.
Ask what the company gives up. An exclusive grant may block other licenses, partnerships, or product plans. It may also limit your own use. Any reserved use should be written down.
A broad exclusive license can come close to a sale in practical effect. Treat it with the same care.
Training rights and retrieval rights.
AI buyers want data for different purposes. The purposes carry different risks. Do not combine them in one grant.
Training rights permit use of the material to train or fine-tune a model. The data influences the model’s parameters. Once that happens, removal can be difficult. The agreement should address what the buyer may retain, what happens at termination, and whether derived models are covered.
Retrieval rights permit a system to access material when answering a query. The content may be quoted, summarized, or cited. This use depends on continued access. It is easier to limit by term and to switch off.
Evaluation rights permit testing and benchmarking only. They are often the narrowest grant. A buyer may need them before deciding on anything larger.
Questions to settle for each right:
- Which purpose is granted, and which are not?
- May the buyer create derivative datasets?
- Can outputs reproduce the licensed content?
- What must the buyer delete at the end of the term, and how is deletion confirmed?
- May the buyer sublicense or give access to affiliates?
Silence in a contract invites dispute. If a right is not granted, say so.
Field-of-use and term limits.
Field of use defines where the buyer may apply the data. Term defines how long.
A field limit might restrict the license to one product, one industry, or one type of model. It might exclude uses that compete with the company’s own business. It might exclude sensitive applications.
A term limit sets an end date. It may include renewal options. A short initial term can give both sides a way to test the relationship before committing further.
Consider these points:
- Does the term cover access, use, or both?
- What survives expiration? Models already trained may be the hardest question.
- Are refreshes or updates included, or priced and scoped separately?
- Can the company terminate for misuse, non-payment, or breach?
- Does the license permit audits or reporting?
Limits protect the company’s other options. They also give the buyer a clear scope, which can shorten negotiation.
What stays with the company.
The agreement should state what the company keeps. Do not rely on assumption.
Typical reservations include:
- Ownership. The company retains title to the licensed material.
- Internal use. The company continues to use the data in its own operations.
- Other licenses. The company may license outside the exclusive scope, if any.
- Excluded material. Customer records, personal information, third-party content, and restricted material stay out.
- Brand and marks. The buyer receives no right to use the company’s name or marks without consent.
- Future data. New material created after the effective date is not covered unless the agreement says so.
- Control of delivery. The company decides the format, the channel, and the access conditions.
Some material should be excluded from the start. Contracts with customers may restrict reuse. Personal information carries legal duties. Contributors may hold rights. Content from third parties may not be yours to license.
A cleaner boundary makes the license easier to sign and easier to defend later.
A licensing checklist.
Use this list before approaching any buyer.
Rights
- Confirm which entity owns the data and can sign.
- Review employee, contractor, and contributor assignments.
- Check customer, supplier, and platform terms for restrictions.
- Identify personal information and confidentiality duties.
- Record any prior licenses, exclusive or otherwise.
Package
- Inventory by source, date range, format, and language.
- Separate usable content from duplicates and empty records.
- Document how the material was collected and processed.
- Prepare a sample that represents the proposed scope.
- Label known gaps and quality limits.
Terms
- Decide exclusive or non-exclusive, and what it covers.
- Define permitted purposes: training, retrieval, evaluation.
- Set field-of-use limits and a term.
- Specify deletion, audit, and reporting duties.
- List what the company reserves.
Approval
- Have counsel review the scope and the agreement.
- Confirm who is authorized to approve disclosure and signature.
- Agree on recipients and access conditions before any sample leaves the company.
If the checklist exposes gaps, fix them first. A gap found by a buyer costs more than one found internally.
When licensing makes sense.
Licensing tends to fit when these conditions hold:
- The company wants to keep operating and keep using the data.
- The data is identifiable and separable from restricted material.
- Rights are documented, or can be documented at reasonable cost.
- The company can describe a field of use it is comfortable with.
- Ongoing delivery or refresh obligations are manageable.
It may not fit when ownership is unclear, when the data is inseparable from customer confidences, or when the business itself is the asset a buyer wants. In those cases, a narrower license, a delay, or a different transaction may be the better course.
Some companies are closing or winding down. Others have inherited archives. Still others hold community content or recorded conversations. Each situation raises its own rights questions. Our pages on acquired archives and company shutdown data cover two of them. More guidance is in Insights.
How representation works.
We act for the seller. We help assess the asset, prepare the package, approach buyers, and compare terms. Your counsel reviews the documents. The authorized owner decides whether to proceed.
The working process describes each stage and the approvals that come before any disclosure. There is no promised outcome or closing date. Timing depends on rights evidence, access to systems, and buyer evaluation.
If you are weighing a license against a sale, start with a confidential seller brief. Describe the data, the entity that holds it, and what you want to keep. We will tell you whether a license looks workable.
Common questions
Before you begin.
Can a company license its data and keep using it?
Yes, if the license is non-exclusive or the exclusivity is narrow. The agreement should state what the company may keep doing with the data. Internal use, customer service, and existing contracts should be reserved in writing.
Is licensing data the same as selling it?
No. A license grants defined permissions for a defined period. A sale transfers ownership of the asset. The contract, not the label, determines which one you have.
Do training rights include retrieval rights?
Not automatically. Training changes a model. Retrieval lets a system look up and show or summarize content at query time. Each right should be granted, limited, or withheld on its own terms.
What if we only have rights to part of our data?
License only that part. Identify the cleared subset, exclude the rest, and describe the boundary accurately. A smaller, well-documented package is easier to defend than a broad one with gaps.