- The agreement grants a defined, non-transferable right to use a copy of the data for model training, and nothing else.
- It prohibits re-identification, resale and attempts to extract the data from trained models.
- It sets delivery, security, deletion, audit and what happens to trained models when the license ends.
- This template is a starting point for discussion, not legal advice. Have a lawyer review any agreement before you sign it.
Clause by clause
1. Definitions
Defines the Data by reference to a schedule (use the data manifest), and defines Model, Permitted Purpose and Derived Data. Precise definitions prevent most later disputes.
2. License grant
Grants a non-exclusive (or exclusive, if agreed), non-transferable, non-sublicensable license to use the Data for the Permitted Purpose during the Term. Everything not granted is reserved. See exclusive vs non-exclusive licenses and sublicensing.
3. Restrictions
The licensee may not re-identify anyone, contact people in the data, resell or share the Data, use it to build a competing dataset, or train models designed to reproduce the Data verbatim. These are the protections that make a license safer than a sale.
4. Delivery and security
The licensor delivers a prepared export, never live system access. The licensee stores the Data in a defined environment with access controls and notifies breaches within a set period.
5. Fees
Fixed fee, installments or royalties. Any broker's success fee is paid by the licensor under a separate agreement.
6. Warranties
The licensor warrants it has the right to grant the license and has applied the agreed de-identification. The licensee warrants it will use the Data only as permitted and comply with applicable law.
7. Indemnities and liability
Each side covers claims arising from its own breach. Liability caps are negotiated. Buyers often ask for a rights indemnity; sellers should cap it.
8. Term and termination
Sets the Term and rights to terminate for breach. On termination, the licensee deletes the Data and certifies deletion. The agreement states whether models trained during the Term may continue to be used, which is the most negotiated clause in AI data deals.
9. Audit
The licensor may audit compliance, usually by written certification with a right to an independent audit on reasonable suspicion of breach.
10. Transparency disclosures
Agrees how the licensee may describe the Data in public training data summaries under laws such as California AB 2013 and the EU AI Act, and whether the licensor is named.
11. General
Governing law, notices, entire agreement and assignment.
Before you use it
- Agree commercial terms first with the license checklist.
- Check your rights and privacy position with the data rights checker.
- Read the full guide to the AI data license agreement.
Read next
Questions
What is the difference between a data use agreement and a data license?
In AI data deals the terms are often used interchangeably. Both grant permission to use data for a defined purpose under conditions. See data use agreement.
Can the buyer keep models trained on my data after the license ends?
Only if the agreement says so. Many licenses allow continued use of models trained during the term but require deletion of the data itself. Negotiate this explicitly.