Permitted use
Permitted use is the purpose and activity boundary of a license. It states what the recipient may do with the covered data. A permission to inspect records for evaluation does not necessarily authorize model training, public redistribution, or another organization’s access. The agreement should make those distinctions understandable.
How it works
Connect the intended activity to named data, users, and outputs. The parties can specify whether processing includes training, fine-tuning, benchmarking, product development, or internal research, and address prohibited uses where appropriate. They may also need to distinguish source records from derived datasets and model artifacts.
Different licenses handle results differently. For example, CDLA—Permissive 2.0 has a separate treatment of computational results. That provision is an example to read, not a rule that automatically applies to a negotiated commercial arrangement.
Why it matters for licensing
A mismatch between a recipient’s technical process and the licensed purpose can undermine a proposed project. Providers and recipients should describe the workflow concretely enough to identify copying, transformation, retention, and access requirements before agreeing to broad labels such as “AI use.”
Example
Fictional example: A recipient receives records to compare two existing models. The proposed permission covers evaluation but does not cover updating model weights. The team must resolve that difference before using the same records in a fine-tuning experiment.
Limitations and misconceptions
A stated purpose alone does not establish lawful collection, remove confidentiality obligations, or authorize third-party content. Enforcement also depends on the agreement and practical oversight. Definitions should be reviewed when a project expands beyond its original purpose.
Questions to ask
- Which steps in the proposed technical workflow are expressly authorized?
- Are derived datasets, model artifacts, and onward sharing addressed?
- What process applies when the recipient wants to change the use?
Sources
Explore whether your business data could be a fit.
Start with a description of your systems—not a data upload.