Which licence questions matter if we publish results or distribute benchmark tasks?
Evaluation buyers should distinguish internal use, model training, hosted access and redistribution when asking which rights a proposed generated release would include.
**Does permission to evaluate include permission to distribute tasks?**
Do not assume that it does. Ask whether the proposed licence covers internal evaluation, training, adaptation and redistribution separately. These are questions for the proposed agreement, not claims about terms MirrorFirm currently offers.
**What if users access tasks through our evaluation service?**
Describe what those users can see and retain. A hosted benchmark that reveals full conversations creates a different distribution pattern from a service that returns only aggregate scores. Include access through an MCP interface in that description where relevant.
**Can we publish examples with our results?**
Ask about publishing task excerpts, complete cases, policy text and derived annotations. Also clarify whether collaborators, contractors or model providers may receive the material as part of the evaluation process.
**Does generated output remove the need for review?**
No. The proposed release is generated material, but buyers should still inspect grounding and disclosure risk and clarify permitted uses. A useful first step is a concise distribution map showing who receives which artefacts and whether they can export them.
Clarify intended downstream uses before embedding generated tasks in a shared benchmark or service.
- Audience:
- Evaluation teams
- Content type:
- FAQ
- Buyer question:
- Scope and getting started
Content release: r-a220eb6e099b243b
Related questions
- How can a buyer adapt a generic sample to professional services billing?
- What integration contract should be agreed before connecting an agent to a pilot environment?
- What should a sample request include so the pilot has a clear exit decision?
- What licence questions should a buyer resolve before using a generated pack?
- What is a useful first pilot for an agent entering a hardship workflow?
- What technical contract should we agree before loading a pilot into our simulator?