How to keep ownership of your company data when you license it
You keep ownership of company data in a license by granting a limited right to use it instead of assigning or selling it. The agreement should state that title stays with you, limit use to AI training, set a defined term, bar resale and free sublicensing, and keep confidentiality and deletion duties. Nothing binds you until you sign.
How do you keep ownership of data you license?
You keep ownership by granting a license rather than making an assignment or a sale. A license gives the buyer a defined right to use copies of your records for a stated purpose and period; title, control and every right you did not grant stay with your company. In practice one sentence does most of the work: the agreement should say plainly that you retain all right, title and interest in the licensed data and that no other rights pass by implication.
That is how SourceX structures company deals. The company keeps ownership, the data is licensed rather than sold, and nothing is binding until the company has agreed price and terms and signed. Deals are typically exclusive for AI training for an agreed term, which is a promise about who else you license to, not a handover of the asset.
This guide covers what to confirm before you negotiate, the clauses that protect ownership and the drafting mistakes that quietly give it away. For the wider trade-offs, read the pros and cons of licensing company data to AI developers.
What do you need in place before you negotiate?
You can only keep, and license, what you actually own or control. Settle four things before a draft agreement arrives.
- A record of what exists. List each system, the years it covers and who can export from it. The metadata-only data inventory builder helps you do that without sharing any records.
- Ownership by source. Under the Copyright Office's Circular 30 on works made for hire, material an employee prepares within the scope of the job generally belongs to the employer. Work by outside contractors usually does not, unless it falls in a listed category with a signed work-for-hire agreement or the rights were assigned in writing.
- Third-party limits. Customer contracts, NDAs, privacy notices and vendor terms can restrict what you may license even when the files sit on your servers. Why data rights determine what a company can license walks through the common cases.
- An authorized signer. The owner, CEO, CFO or another authorized representative must be able to approve the scope and sign.
How to license data without giving it away, step by step
Work through the draft in this order. Each step maps to a clause you can point to.
- Read the grant first. Find the operative sentence. "Grants a license" keeps title with you; "assigns", "sells", "conveys" or "transfers all right, title and interest" moves it. If the document is titled as a purchase agreement, ask for a license structure before negotiating anything else.
- Add a retained-ownership sentence. State that you keep all right, title and interest in the licensed data, that the buyer receives only the rights expressly granted, and that nothing else passes by implication.
- Define the field of use. Limit use to training and evaluating machine learning models. Exclude resale, publication, re-licensing of the raw records and any use to identify or contact the people or businesses that appear in them.
- Set the term and the scope of exclusivity. Name the start date, the length and what happens at the end. If the license is exclusive, define exclusivity as a promise not to license the same dataset to other AI developers for training during the term, so your own internal use stays untouched.
- Restrict transfer. Allow no sublicensing except to service providers working for the buyer under equally strict terms, and no assignment of the agreement without your consent, including on the buyer's change of control.
- Keep confidentiality and security duties. Treat the dataset as your confidential information, require reasonable security and agree de-identification and redaction rules before any preparation work starts.
- Write the exit. Require return or deletion of data copies at the end of the term, with written confirmation, and say explicitly how copies, derived datasets and models trained during the term are treated. The companion page on what happens to data when an AI training license ends goes deeper on this clause.
A short grant clause that combines steps 1 to 5 might read as follows. It is Illustrative wording for discussion with your counsel, not a SourceX contract term.
Why does exclusivity need careful drafting?
Because an exclusive grant can carry more than its label suggests. Section 201 of the Copyright Act lets ownership be transferred in whole or in part, and lets any single exclusive right be transferred and owned separately. That divisibility is what makes a narrow grant possible, and it is also why a broadly worded exclusive grant can move more than you meant to.
Ask your counsel how the exclusivity language will be characterized, keep it tied to one field (AI training) and one term, and confirm you can still use your own records for operations, analytics and your own internal AI tools. This is general information, not legal, tax or financial advice. Confirm with your own counsel, tax adviser or professional body before acting.
Price, term length and exclusivity trade against each other, so settle ownership language first and then turn to how to negotiate an AI data licensing deal.
Which drafting mistakes give ownership away?
Most ownership problems come from a few loose phrases rather than from the deal itself.
| Mistake | Why it hurts | Fix |
|---|---|---|
| Accepting a "data purchase" or asset-sale structure | Title moves to the buyer, and you may lose the right to use or license the records again | Insist on a license grant plus a retained-ownership sentence |
| Purpose written as "any lawful purpose" | The buyer can go beyond training, including resale or publication | Limit use to training and evaluating models |
| "Perpetual and irrevocable" with no carve-outs | There is no end date and no exit if the buyer breaches | Set a defined term and termination rights for breach |
| Free sublicensing or assignment | Copies can reach parties you never approved | Allow only bound service providers and require consent to assign |
| Exclusivity left undefined | It may be read to restrict your own internal or non-AI uses | Tie exclusivity to third-party AI-training licenses only |
| Licensing material you do not own | You breach your own warranties and may owe indemnity | Sort records by source and exclude contractor or client material without rights |
| No deletion mechanics | Copies linger after the deal ends | Require return or deletion with written confirmation |
Illustrative example: a survey firm keeps its archive
Illustrative: Northfield Survey Group, a fictional 160-person engineering survey firm, receives a first draft that calls the arrangement a "dataset acquisition" and says the firm "transfers all right, title and interest" in twelve years of project correspondence and field reports. The owner's counsel replaces the transfer with a limited license for model training over a fixed term, adds a retained-ownership sentence and restricts exclusivity to other AI-training licenses.
The inventory also shows that a block of older field reports was written by subcontracted crews whose agreements contain no assignment clause, so those reports are carved out. The firm signs only after redaction rules for client names are agreed, keeps full use of its archive for its own estimating work and receives one all-in payment for the licensed scope.
How does SourceX protect ownership during the process?
The process is built so that you decide what leaves the company and on what terms.
- You complete an inventory of systems and years, not an upload of records.
- Price and terms are agreed with you before buyers review the opportunity.
- De-identification and redaction requirements are agreed with you before any work begins.
- Data is delivered only after an executed agreement and your authorization.
- You receive one all-in price, with SourceX's fee included and no separate charges, as a one-time payment typically within about 60 days of invoicing once the buyer selects the data.
If an advisor introduced you, their referral reward is a share of SourceX's fee and is never deducted from what your company receives. How SourceX referrals work covers each stage. If a sale of the business is on the horizon, read whether licensing data affects company valuation before you sign.
A final ownership check before you sign
Run this list against the execution version, not an earlier draft.
- The operative clause says license, and nothing says assign, sell or convey.
- A sentence confirms you retain all right, title and interest.
- Use is limited to training and evaluating models.
- The term has an end date, and exclusivity covers only AI-training licenses.
- Sublicensing and assignment need your consent.
- Redaction and de-identification rules are written down.
- Return or deletion of copies is required at the end, with confirmation.
- Your own counsel has reviewed the final text.
Next step
If you own a company with years of records and want to license them without giving them up, apply directly at sourcex.si/apply. If you advise business owners and want to make introductions, register as a partner. Partners earn 25% of the eligible platform fees SourceX actually collects from the referred company's licensing deals, capped at $100,000 per referred company, and rewards become payable only after the buyer pays and SourceX receives its fee.
- Step 1Share your linkSend your personal link to a company you know.
- Step 2Company appliesThe company applies itself at /apply.
- Step 3Buyer selects and paysThe buyer selects and pays for the data and SourceX receives its fee.
- Step 4You get your rewardYour share of SourceX fees becomes payable.
Common questions
Is an exclusive data license the same as selling the data?
No. A sale or assignment moves title to the buyer. An exclusive license keeps title with your company and promises that, for an agreed term and field, you will not license the same records to anyone else. Because exclusive rights can be divided and transferred separately under copyright law, the wording still matters, so have counsel confirm the grant is limited to AI training and a defined term.
Can we keep using our own records during an exclusive AI training license?
You should be able to, and the agreement should say so. Define exclusivity as a restriction on licensing the same dataset to other AI developers for training during the term, and add a sentence confirming your company keeps full use of its records for operations, reporting, analytics and its own internal tools. If a draft is silent on this, ask for that sentence before signing.
What happens to a model trained on our data after the license ends?
Deleting copies of the dataset does not by itself remove what a model has already learned from it, so treat this as a separate question in the agreement. Ask the draft to state how data copies, derived datasets and models trained during the term are handled at the end, and have counsel confirm the wording matches what you expect before you sign.
Who owns documents written by contractors rather than employees?
Often not the company. Under US copyright rules, work an employee creates within the job generally belongs to the employer, but a contractor's work usually stays with the contractor unless it falls in a listed category covered by a signed work-for-hire agreement or the rights were assigned in writing. Check contractor agreements before including that material, and leave it out when rights are unclear.
Does SourceX take ownership of our data?
No. SourceX manages the licensing process, from inventory and rights review to buyer review, contracting and delivery, but the company keeps ownership and the data is licensed, not sold. Nothing is binding until the company agrees price and terms and signs, and data is delivered only after an executed agreement and the company's authorization.
Should a lawyer review the license even if the terms look standard?
Yes. Small wording changes, such as transfer instead of license or any purpose instead of model training, change what the buyer receives. A lawyer who handles technology or intellectual property agreements can check the grant, exclusivity, warranties, indemnities and exit terms against your own customer contracts and policies, which a template cannot do for you.
Related pages
- Pros and cons of selling or licensing company data to AI developers
- Build a metadata-only business data inventory
- Why data rights determine what a company can license
- What happens to your data when an AI training license ends?
- How to negotiate an AI data licensing deal: priorities in order
- How SourceX US company data referrals work
Free resources
- Portfolio data opportunity scanner — Screen several companies in one session.
- Working capital calculator — Net working capital, current ratio and quick ratio.
- Due diligence checklist generator — A tailored document request list by deal type.
- All free tools · MCP resource center
By SourceX Partnerships Team · Published 2026-10-09 · Updated 2026-10-09
Know a US company with valuable proprietary data?
Become a referral partner from anywhere we support, get your link and introduce an owner or authorized decision-maker.
Refer a company →I own a business
Explore licensing your company's data to AI developers worldwide. Start a short assessment; no uploads needed.
Start an assessment