Your terms travel with every passage.
TRIVDA makes your archive answerable by AI systems without taking it out of your hands. You set the permitted uses, they are checked on every request, and you earn each time your data answers one.
- Permitted uses
- set per dataset
- Training
- only if you grant it
- Withdrawal
- at any time
From a folder to a licensed dataset.
-
01
You bring the material
Upload files, or connect the store they already live in: Google Drive, SharePoint, S3 and the other major providers.
-
02
We make it machine-readable
Each file is parsed, cleaned and split into passages: the short extracts an AI system retrieves and cites. Scans are read by OCR, recordings are transcribed, and personal data is screened out.
-
03
You state the terms
Who owns it, under what licence, in which jurisdiction, and what it may be used for. That record is what every later request is judged against.
-
04
It answers, and it earns
Once offered, your dataset can answer questions from any TRIVDA customer. It earns a share of every paid answer it contributes to.
What is on record before anything is searchable.
- Read at query time
- Every value here is checked when a question arrives. Permitted uses decide whether a passage may be returned.
- No warranty, no publication
- A dataset without a warranty on file cannot be offered at all. Without a declared licence it is not even indexed.
- Readiness is earned
- Bronze, silver or gold, from how cleanly the source parsed and how it came through screening. Below the publishing floor it stays out of the index.
- Owner
- Acme Industrial AB
- Warranty
- On fileYour own statement that you are entitled to license this material. Without it, nothing is published.
- Licence basis
- The owner’s own terms
- Jurisdiction
- EU
- Permitted uses
- search, retrieval
- Training
- Not granted
- Offered
- To every TRIVDA customerOffering it means every TRIVDA customer can reach it. Keep it internal and nobody outside your organisation can.
- Readiness
- Gold
You earn when your data answers.
- You are paid
- per query answered
- Measured by
- passages returned
- Written
- with the charge
- Recorded
- per dataset, per query
When a customer pays for a query, the fee is divided between the datasets that answered it, in proportion to the passages each one supplied. Commercial terms are agreed with you before anything is published.
Four things that cannot happen.
-
An internal dataset cannot be offered
A dataset you keep internal is reachable only by your own organisation. Access is refused by default and opened only by your decision.
-
Search cannot become training
Uses you have not granted are denied outright. A request to train on a dataset licensed for search alone returns nothing.
-
Unrighted material cannot be indexed
A file that arrives with no declared licence is quarantined instead of indexed.
-
Your data cannot leave for a third party
Parsing, reading scans and transcribing recordings all run on our own infrastructure. Your material is not sent to an outside service to be processed.
| What you do | What stops | What you keep | Reversible |
|---|---|---|---|
| Withdraw | It stops answering, for everyone. | The data, and everything it has earned. | Yes |
| Revoke one organisation | That customer alone loses access. | Every other customer carries on. | Yes |
| Change the permitted uses | Any use you remove is refused from the next request. | The dataset stays where it is. | Yes |
| Erase | The passages are deleted from the index. | The record of what it earned. | No. The identifier stays suppressed. |
Tell us what is in your archive.
Roughly what it is, and what you would allow it to be used for.