How is generated conditional-approval material licensed, and what limits apply?

**How is the material licensed?**

Licence terms are agreed for each engagement. Ask what uses are covered, such as internal training, evaluation, or redistribution inside a product, and get the answer in writing before you build on the data.

**Can I train a model on it?**

That depends on the licence you agree. State your intended use up front, including whether the output would shape a model you later sell or share.

**Can I publish evaluation results?**

Raise this early. Whether you may publish scores, examples or task text is a licensing question, not a technical one.

**Is the data real?**

No. The commercial product is generated, synthetic material. It is built to reflect how conditional approvals work in practice, but it is not a copy of any organisation's records, and it carries no assurance of anonymity or compliance on its own. Inspect it for grounding and disclosure risk, and run your own review where your obligations require it.

**What limits should I expect?**

Generated content approximates a process. It will not match your internal field names, edge cases or volumes unless you scope them, and some behaviours will only appear in your live systems.

**What should I ask for first?**

A small sample in your target format, with the schema, policies, tasks and ground truth, so your team can review structure before you discuss anything larger.

**Who decides if the sample is good enough?**

You do, against success criteria written before the sample arrives.