Selling data to AI companies: score your readiness before you pitch
Most dataset conversations stall on one question the seller cannot answer cleanly. This scorecard turns the usual diligence questions into seven factors you can grade yourself on, red, amber or green, before you contact anyone.
Seven checks, three questions
- May you license it? Chain of title and contract freedom.
- Is it safe to share? Personal information, sensitive categories, third-party works.
- Can a buyer evaluate it? Documentation and uniqueness.
Original source
Who created or collected it?
Rights chain
What contracts and permissions apply?
Proposed AI use
Does the allowed scope cover this use?
Check copyright, contract restrictions and personal-information obligations separately. Access alone is not permission to license.
The seven-factor scorecard
| Factor | Green looks like | Red looks like |
|---|---|---|
| 1. Chain of title | You created the data or hold written assignments | Data sits on your servers but came from customers or vendors |
| 2. Contract freedom | No customer/vendor terms restrict reuse | Terms of service promise data is used only to provide the service |
| 3. Personal information | None, or de-identified with a documented method | Unknown; free-text fields never reviewed |
| 4. Sensitive categories | Excluded or separately consented | Health, biometric or children's data mixed in |
| 5. Third-party works | Inventoried, licensed or removed | Scraped articles, stock images, music beds |
| 6. Documentation | Datasheet + provenance dossier exist | Only a folder name and an estimated row count |
| 7. Uniqueness | Not publicly crawlable; hard to recreate | Mirrors content already on the open web |
How to read your score
Any red in factors 1–4 is a stop sign, not a negotiating point: fix it, carve the affected records out, or don't offer that set. Reds in 5–7 reduce what the data is useful for, but they rarely end the conversation if disclosed honestly.
A green scorecard does not mean someone will buy. It means you can answer the questions a serious counterparty will ask without improvising.
The FTC angle many sellers miss
The Federal Trade Commission has warned that companies which quietly change their privacy commitments to repurpose user data for AI training may be acting unfairly or deceptively. If your data came from your own users, read what you told them at collection before scoring factor 2 green.
Move from readiness to a commercial brief
Define a useful subset rather than offering everything in your archive. Describe the model task it could support, the permissions you can document and the delivery work you can commit to. A licensing conversation is about that specific package—not a price for every record your business has ever collected.
How this affects value
Every red factor you resolve before outreach removes a reason for a buyer to discount, narrow or delay the deal.
See what shapes a dataset's value