Skip to content
MirrorFirm

What licence and usage questions should a lab raise before starting?

A lab should ask what the licence permits for training, evaluation and redistribution, how versions are identified, and what limits apply to derived models and published results.

**What should be settled in writing?**

Terms are agreed with MirrorFirm, so none are assumed here. Raise these points early.

- **Permitted use:** training, fine-tuning, evaluation or all three, and whether internal and external model releases are covered. - **Derived works:** whether models, embeddings or summaries trained on the material carry any restriction. - **Redistribution:** whether you may share tasks or traces with partners, contractors or in published papers. - **Evaluation reporting:** whether results from the task packs can be published, and in what form. - **Versioning:** how releases are identified, so you can cite the exact version used in a run. - **Updates and withdrawal:** what happens if a record is found to resemble a real entity too closely.

**Does the licence settle disclosure risk?**

No. A licence sets permitted use. It does not remove the need to inspect generated output for grounding and disclosure risk, and it does not stand in for your own legal or privacy review.

**Who should be involved?**

Bring someone from legal or procurement alongside the technical lead, so the sample request and the licence discussion move together instead of one blocking the other.

Raising licence questions at the sample stage avoids discovering restrictions after your team has built pipelines around the data.

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

Content release: r-a220eb6e099b243b

More on Account setup