Prospective supplier review
Make your human dataset easier for serious AI buyers to evaluate and license.
NaraDatum reviews organisations that hold or can lawfully license human data. This form is how you ask to be considered for that review.
It asks who you are, what you hold and in which languages. Provenance, rights, consent, metadata, quality control and intended-use fit are reviewed on each dataset separately. No dataset is treated as supply-ready until the review is complete and the required licensing authority, availability and transaction terms can be established.
What the process does for a dataset that holds up
Buyer review can stop when rights cannot be traced, consent cannot be evidenced, or intended-use fit cannot be tested. This process records the evidence and gaps before a commercial discussion. It is not a promise of demand or a purchase.
Qualified requirements, not enquiries
Buyers submit a structured requirement: use case, languages, speaker structure, volume and timeline. Your data is matched against a stated brief rather than a browse.
A structured Dataset Passport
What was reviewed, what was established and what was not, in one document a buyer’s legal and technical reviewers can read without a call. See an illustrative Dataset Passport.
Controlled sample access
Evaluation material is shared under agreed conditions with a named recipient, rather than posted to a catalogue where anyone can take a copy.
Rights and provenance evidence, recorded clearly
Chain of title, permitted use and consent scope are reviewed in a private process and the evidence position is recorded for the relevant dataset. Review does not create, transfer or expand licensing rights, and a buyer may still conduct its own diligence.
Buyer-facing evaluation
Fit is assessed against the buyer’s actual use case (ASR, voice agents, TTS, benchmarking), so a set is judged on the job it is for, not on a single quality score.
Commercial protection and managed transaction
Terms, permitted use and delivery controls are handled as part of the transaction. Your rights position is not weakened to close a deal, and a buyer licence never exceeds what you actually hold.
What this application is, and what it is not
Applying is not verification.
Submitting this form records an application. It does not verify your organisation, your claims, or your data.
Supplier approval does not verify every dataset.
An approved supplier is an organisation we are willing to review datasets from. Each dataset is still reviewed on its own before it can be used.
Declared inventory is not verified inventory.
Everything you write below is a declaration by you. Until we have reviewed evidence, it stays a declaration and is treated as one internally.
NaraDatum reviews provenance, rights, consent, metadata, quality control and intended-use fit before any dataset is considered.
A material gap in any dimension can prevent a dataset from being considered. The review records what evidence supports each conclusion and what remains unresolved.
This application creates no obligation on either side.
It is not a listing, not an approval, not an offer, and not a commitment to buy, license or engage commercially.
Who should apply
A good fit
- You own human speech, conversational or multimodal data, or you hold documented authority to license it.
- You can explain where the data came from and how the people in it agreed to its use.
- You can produce contracts, consent records or release evidence later, under a controlled private process.
- You run, or are willing to run, a defined quality process over what you deliver.
Supplier-owned data remains supplier-owned unless a written agreement says otherwise.
Not a fit
- Data scraped or harvested without a lawful basis for the use you are proposing.
- Data where the chain of rights cannot be traced back to a source you can name.
- Synthetic or model-generated material presented as human data.
- Individual contributors offering their own voice. That is a different route. See applying to contribute.
What NaraDatum reviews
Provenance
Where the data actually came from, through every hand it passed, and whether that chain can be evidenced rather than asserted.
Rights
Who holds them, what they permit, and whether they extend to commercial AI training and to the licence a buyer would need.
Consent
What the people recorded were told, what they agreed to, and whether that agreement covers the intended use.
Metadata
What is labelled, how consistently, and whether the labels mean the same thing across the dataset.
Quality control
How defects are found and corrected before delivery, and what happens to material that fails.
Intended-use fit
Whether what you hold fits the use a buyer has stated, and whether the evidence supports that use.
What happens next
Applications move through the stages below. Not every application reaches every stage, and progressing through one stage does not imply the next.
- 01Application receivedYour declarations are recorded. Nothing is verified at this point.
- 02Initial fit reviewWe read what you sent against current and expected buyer requirements. This is an internal fit assessment, not proof of demand.
- 03Information requestIf there is potential fit, we ask specific follow-up questions in writing.
- 04Controlled evidence reviewContracts, consent records, identity documents and sample material are exchanged through a private, agreed process, never through this website.
- 05Dataset-specific reviewEach dataset is assessed on its own provenance, rights, consent, metadata, quality control and intended-use fit.
- 06Commercial discussion where there is real fitTerms may be discussed where a reviewed dataset could fit a specific buyer requirement. Buyer interest is not a purchase commitment or proof of wider demand.
- 07Approval, conditional approval, or declineConditional approval means named conditions must be met before anything proceeds.
Do not send evidence through this form
This form accepts typed answers only. There is no file upload on this page, and there is no way to attach anything to it.
Contracts, consent records, identity documents, audio, transcripts and datasets are handled later, in a controlled private process that we agree with you first. Please do not paste them here, and please do not include links to them.
Describe what exists in words. That is enough for this stage.
The application
Nine short steps. Seven questions about your organisation and your data are required, as are the four acknowledgements near the end. Everything else is optional at this stage.
Application received
Application received
Your declarations have been recorded. Nothing has been verified, and this is not an approval.
If your application reaches initial fit review and there is potential fit, we will write to the email address you gave with specific follow-up questions.
This form does not send you a confirmation, and there is no fixed response time. If you need to reach us in the meantime, email hello@naradatum.com.