Which licence questions should a model lab resolve before using a release?

**Does access settle every downstream use?**

A buyer should not assume that it does. Ask which uses the proposed licence would permit, including pretraining, fine-tuning, evaluation, internal research and use in commercially deployed systems. These are questions to resolve, not statements of MirrorFirm's terms.

**What about derived artefacts?**

Clarify how the proposed agreement treats transformed datasets, model-generated responses, evaluation traces and trained model weights. A lab may handle each differently, so a broad reference to derived material can leave practical uncertainty.

**Can collaborators receive the material?**

Ask about access by contractors, affiliated entities and external research collaborators. Specify whether they need full records, task instructions or only aggregate findings.

**Can results be published?**

Distinguish publishing evaluation findings from redistributing prompts, answers or complete environment assets. A public paper or downloadable demonstration may expose more material than an internal experiment.

**What is the useful first step?**

Prepare a short intended-use statement covering users, systems and planned outputs. MirrorFirm and the buyer can then discuss licensing against concrete activities without inventing commercial terms or assuming permissions from the delivery format.