Not legal adviceThis is a checklist to organize your review. Have qualified counsel review any agreement.
The checklist
| # | Term | What good looks like |
|---|---|---|
| 1 | Parties | Correct legal entities on both sides |
| 2 | Ownership | You keep ownership; buyer gets a license only |
| 3 | Scope | Systems, teams, date ranges and file types listed in a schedule |
| 4 | Exclusions | Written list of what's off-limits |
| 5 | Permitted use | Training and evaluation of models, stated plainly |
| 6 | Third parties | Whether data or derivatives can be shared or resold |
| 7 | De-identification | Who removes PII, method, and verification |
| 8 | Sample review | Your right to approve samples before use |
| 9 | Security | Access limits, isolation, device and staff controls |
| 10 | Retention | Fixed period |
| 11 | Deletion | On request and at term end, with written confirmation |
| 12 | Exclusivity | Non-exclusive unless you're paid for exclusivity |
| 13 | Payment | Amount, milestones, timing and what triggers each payment |
| 14 | Warranties | Reasonable promises about rights; not unlimited |
| 15 | Liability and breach | Caps, notice timelines and who covers what |
Red flags
- "Perpetual, irrevocable" use with no deletion
- Ownership transfer hidden in the definitions
- No right to review samples
- You carry all responsibility for PII removal
- Unlimited warranties or indemnities from you
Related: is it safe to sell company data to AI?
Frequently asked questions
What should be in an AI data licensing agreement?
Scope, exclusions, permitted use, third-party limits, de-identification, sample review, security, retention, deletion, exclusivity, payment terms, warranties and liability.
What are red flags in a data licensing contract?
Perpetual use with no deletion, hidden ownership transfer, no sample review, all PII responsibility on you, and unlimited warranties.