documented review,
not legal guarantee.
Our review is a documented assessment for a defined intended use. It is not a guarantee that a dataset is legally perfect for every possible scenario — and we'd rather be honest about that than dress it up.
what each layer checks. against what evidence.
Same structure every time. Reviewers, providers and buyers all read the dossier in the same shape — the only thing that changes is which section they care about most.
01
Technical review
Whether the dataset is derived from real data, which generation method was used, whether there are limitations in source data, and whether key generation parameters are recorded.
- Origin — derived from real data, fully synthetic, or hybrid (and how).
- Generation method — model, technique, version, seed dependencies.
- Source data limitations — gaps, biases or coverage holes inherited from the seed.
- Parameter record — which parameters were locked, which were swept, and what was logged.
02
Legal and licensing review
Legal status, relevant regimes, rights to the dataset, permitted uses, prohibited uses, buyer responsibility, provider responsibility.
- Legal status — basis on which the dataset can be offered.
- Relevant regimes — the legal context the dataset is offered under (jurisdiction specifics live in the licence).
- Rights — provider's rights to the underlying material.
- Permitted / prohibited uses — written out in the licence, not implied.
- Responsibility split — what stays with the buyer and what stays with the provider.
03
Ethical review
Sensitivity of scenarios, potential for misuse, risk to specific groups, and clarity of guidance for reviewers.
- Scenario sensitivity — categories of content that warrant reviewer notes.
- Misuse potential — uses the dataset enables that the licence must therefore forbid.
- Group-level risk — disparate impact on specific groups, where applicable.
- Reviewer guidance — clarity of guidance the dossier carries forward to downstream reviewers.
what verification is. what it isn't.
Is
- A documented assessment for a defined intended use.
- A consistent structure across every listed dataset, so reviewers compare like with like.
- A written record of what we checked, what we didn't, and what evidence backed each conclusion.
Isn't
- A guarantee that a dataset is legally flawless for every possible scenario.
- A substitute for your own legal, security or model-quality review.
- A static stamp — outcomes are revisited as the dataset evolves, and the dossier carries the version trail.
read a real dossier to see the standard in practice.
The standard is most legible inside an actual listing — open one and follow the references back to this page.