What is chain of title for AI training data?
Chain of title for training data is the documented trail showing who created a dataset, who owns it now and what permissions allow it to be licensed. AI buyers use it to confirm a company can grant the rights it offers, which is why ownership is the first screening question.
What is chain of title for training data?
Chain of title is the documented trail showing who created a dataset, who owns it now, and what permissions allow it to be licensed. A buyer reads it the way a title company reads a property record: if a link is missing, the asset is hard to transfer or license. This is general information, not legal, tax or financial advice.
For a referral partner, it is the reason "who owns it?" is the first screening question. SourceX reviews rights with the company, but you can spot a weak chain in a ten-minute conversation.
How the chain is built, link by link
A chain has four links. Each needs paper behind it.
- Creation: who produced the records. Employee work is generally owned by the employer, while contractor work may need a signed assignment. The Copyright Office's circular on works made for hire explains that an employer is the author of work an employee prepares within the scope of employment, and that commissioned work counts only in listed categories with a signed written agreement. The statutory definition sits in 17 U.S.C. section 101.
- Ownership: who holds the rights today, including after acquisitions, mergers or asset sales.
- Permissions: what customers, employees and vendors allowed in contracts, notices and policies.
- Transfer: whether the company can grant the license it plans to grant, for example an exclusive term for AI training.
Under 17 U.S.C. section 201, ownership can be transferred in whole or in part, and exclusive rights can be held separately. That is why a company can license specific rights while keeping others.
Documents that prove each link
| Link | Documents a buyer expects | Common gap |
|---|---|---|
| Creation | Employment agreements, invention and IP assignment clauses, contractor agreements | Freelancers with no written assignment |
| Ownership | Corporate records, asset purchase agreements, merger documents | Acquired company archives with unclear transfer |
| Permissions | Customer contracts, privacy notices, employee handbooks, call-recording notices | Contracts that reserve customer data to the customer |
| Transfer | Authorization from owner, CEO, CFO or other sponsor; board consent where required | Nobody with authority has been asked |
None of this is something a partner collects. The company assembles it with SourceX during rights review, and the data inventory builder is a place to list systems before that conversation.
Chain of title vs similar terms
| Term | What it answers | Where it differs |
|---|---|---|
| Chain of title | Who created and owns the data, and who may license it | Looks backward through history |
| Chain of custody | Who handled the data and how it was kept intact | Concerns handling, not ownership |
| Provenance | Where the data came from | Broader; includes origin and processing steps |
| Data lineage | How data moved between systems | Technical flow, not legal rights |
Buyers care about all four, but chain of title is the one that speaks to whether the company can grant the rights it offers.
An illustrative example
Illustrative: a fictional 120-person logistics software firm wants to license ten years of support tickets. Its employees wrote the replies, so creation looks clean. But some tickets contain pasted customer documents, and two large customer contracts say customer content stays customer property. The firm's chain is solid for its own replies and weak for the pasted content, so the scope would be narrowed or the pasted material removed before any buyer sees it.
Why it matters for referral partners
A weak chain of title can stop an otherwise promising company. Records that mainly belong to clients, such as an agency's or outsourcer's work product, do not qualify without client consent. The AI data intermediary explainer describes how a platform handles this review so the owner does not have to, and the guide to enterprise licensing deals shows where rights fit in the deal.
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 only after the buyer pays and SourceX receives its fee. No reward is guaranteed. Confirm with your own counsel or professional body before acting on any legal point here.
Related reading
Chain of title becomes concrete in negotiation threads as training data, where counterparties' words appear in the records, and in how much data frontier models use, which explains why buyers insist on clean rights at scale.
Next step
Ask one owner three questions: did your company create these records, do customer contracts allow licensing, and who can authorize it? Run the company fit checker for a preliminary screen, then register as a partner and introduce them. See how the process works first if you want the full sequence.
- 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
Who owns emails and documents employees write?
Generally the employer owns work an employee prepares within the scope of employment, under the work made for hire rules. Contractors are different: their work needs a signed assignment or qualifying written agreement. Counsel should confirm for a specific company, because contracts and state law can change the answer.
Does a company need written proof for every record?
Not for every record. Rights review usually works at the level of systems and categories, such as employment and contractor agreements, customer contracts and policies. Gaps are typically handled by narrowing the scope or excluding the affected material, and counsel can advise on what a specific buyer will want to see.
What if customers own part of the data?
That portion is out of scope unless customers consent in writing. Many companies still hold a large body of their own records, such as internal documents, planning, engineering and operations, that can be licensed without customer content.
Can an acquired or wound-down company still show chain of title?
Yes, if the acquisition or wind-down documents show who holds the assets and the data still exists. A court, trustee or assignee that controls assets must be involved before any license. Status alone does not disqualify a company.
Do partners need to collect chain of title documents?
No. Partners make introductions and share basic fit information only. The company and SourceX handle rights review, and partners never export, upload or describe confidential records.
Related pages
- Build a metadata-only business data inventory
- What is an AI data intermediary?
- Enterprise AI data licensing deals: what advisors should know beyond the headlines
- Negotiation threads as AI agent training data
- How much data are frontier AI models trained on, and does size matter?
- Check Company Fit for Data Licensing
Free resources
- Referral earnings calculator — Hypothetical partner earnings with the per-company cap.
- Cash conversion cycle calculator — DIO, DSO, DPO and the cash conversion cycle.
- Operational data inventory builder — List systems, record types, years held and owners.
- 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