What licence limits and disclosure questions should we raise before starting?

**Do licence terms come with the sample?**

Terms are agreed for each engagement. Ask for them in writing before you use a sample beyond internal inspection.

**What should we ask about use?**

- May the material be used for training, evaluation or both? - May it be used to fine-tune a model, and may that model be shared? - Can results be published, and can example cases appear in papers or documentation? - May contractors or other teams access it?

**What should we ask about redistribution?**

- Can you include cases in a public benchmark, or only use them internally? - What happens to derived scores and labels?

**What should we ask about the data itself?**

- It is generated, synthetic material. It is not real business records. - Ask how it was grounded in observation of business processes. - Ask what disclosure checks apply, and run your own: look for identifiers or patterns that resemble real entities. - Generated does not mean anonymity, privacy compliance or regulatory approval is assured. Treat those as questions to test, not assumptions.

**What if we need something not covered?**

Say so early. Limits you discover after building on a sample are harder to fix than limits you agree at the start.