What licence questions and limits should we raise before starting?

**What can we license?** A buyer can ask for a dataset, a task and evaluation pack, a simulated environment, or a combination. The unit, volume and term are agreed in scoping; this answer does not set them.

**What should a licence cover?** Ask about permitted uses (training, evaluation, internal benchmarks), whether results or derived benchmarks can be published, whether the data can be shared with contractors, and what happens to derived models if the licence ends. Get answers in writing.

**Are the records real?** No. Everything delivered is generated. It is informed by direct observation of how real businesses operate, but it is not a set of real business records.

**Does generated data mean no privacy risk?** No. Do not assume it. Generated text can still resemble real documents or identifiers. Inspect samples for grounding and disclosure risk, and run your own review before any use that carries obligations.

**Does it carry regulatory approval?** No. Nothing about the data implies compliance or approval for any use. Your own legal and risk teams decide that.

**Can we request custom requirement tables?** Scope it. You can supply your own policy or ask for a table to be drafted for your review. Say which one governs when the answer key is built.

**What limits should we plan for?** Domain coverage, document formats and environment fidelity are set by the pilot scope. Ask what is out of scope early.