Skip to content
MirrorFirm

Which licence questions should a model lab settle before using a generated pack?

A model lab should clarify permitted training, evaluation and redistribution uses, including access through shared infrastructure, without assuming that synthetic status determines usage rights.

**Does generated material automatically permit every training use?**

No. Generated status describes the material, not the permissions attached to it. A buyer should describe the intended use and ask for the applicable licence scope before ingesting a pack.

**Which uses should the lab distinguish?**

Ask separately about model training, internal evaluation, hosted evaluation services and publication of task examples. If affiliates, contractors or external researchers will access the material, include those access arrangements in the discussion. Shared storage can create distribution questions even when the original buyer remains responsible for the experiment.

**What about transformed material and model outputs?**

Clarify how the licence treats edited tasks, combined datasets, derived annotations and excerpts reproduced in reports. Ask whether any conditions affect checkpoint sharing or downstream access; do not infer the answer from permission to run internal training.

**What should accompany a sample request?**

Provide a short use description covering users, infrastructure, transformations and intended publication. Ask which permissions apply to the sample and which would need to be settled for a broader release. These are questions for scoping an agreement, not a statement of MirrorFirm's current contract terms.

Turn intended uses into explicit licence questions before material enters shared training or evaluation infrastructure.

Audience:
Model labs
Content type:
FAQ
Buyer question:
Scope and getting started

Content release: r-a220eb6e099b243b

More on Policy change propagation