What licensing questions and limits should I raise before starting?

**What licensing questions should I raise?** Raise them at the start. Terms are agreed for the engagement, so ask in writing what uses are covered.

**Which uses should I name?** Name the ones you plan: training, fine-tuning, evaluation only, internal benchmarking, publishing results, and sharing with partners or customers. Ask whether a model trained on the data can be released, and whether the evaluation tasks themselves can be published.

**Are there limits on what the data is?** Yes. It is generated, not a copy of any business's records. It is informed by direct observation of how businesses work, but it is not a source of real transactions or real parties.

**Does generated data remove privacy or compliance duties?** No. Do not treat it as evidence of anonymity, privacy compliance or regulatory approval. Inspect the sample for grounding and disclosure risk, and apply your own review before using it in a regulated setting.

**Can I hold back a test set?** A buyer can ask for separate training and held-out sets, split by decision so related records do not cross. Discuss whether the held-out set is kept out of any wider release, so it stays useful for testing.

**Who sets scale, price and timing?** Those are agreed in the engagement. This page does not set them, so ask for them in writing before you commit.