Licensed engineering data for AI labs, with the rights and the lineage attached.

GenSet Data is building a way for companies to license the engineering files they own to AI labs: preserved exactly as received, and delivered as versioned, signed dataset releases that engineers can load and legal teams can accept.

Coming soon. We are looking for the first supplier and the first AI lab to pilot it with. Write to us: team@gensetdata.com

How it will work

  1. Agreement A supplier signs an agreement naming what is licensed, to whom and for what.
  2. Intake Files arrive with checksums, and their rights are recorded before any processing.
  3. Processing Versioned plugins build the dataset and record every step.
  4. Release Two people approve a frozen dataset version, released with a signed manifest.
  5. Delivery The release is copied to the lab, which verifies every checksum and the signature.

How it is designed

Provenance

Every file will arrive under a signed agreement and a manifest the supplier writes: path, SHA-256 checksum, declared file type and a declaration of third-party content. We will recompute the checksum and confirm the file type before anything is cataloged. Rights evidence and an export-control classification will be recorded for each file, and no processing will run until both are in place.

Lineage

Raw files will never be modified. Every processing step will be a versioned plugin run as a job that records what it read, what it wrote and which version did it. A release will be a frozen list of records with a signed manifest that names every object and its checksum, and rebuilding the same dataset version will give the same manifest.

Isolation

Each supplier and each lab will be its own organization, and every query will be scoped to one. Uploads and downloads will use short-lived credentials limited to one storage prefix, and file bytes will never pass through the application. A request for another organization's record will be answered as not found.

For suppliers

For design firms and manufacturers that own their engineering files and want to license them on their terms.

  • Nothing will move before a signed agreement that names what is licensed, to whom and for what.
  • You will hand over a manifest with checksums, ownership evidence and a list of third-party components. Our command-line tool will check the manifest before a byte is transferred, resume after a dropped connection and verify every file on arrival.
  • Your files will sit in their own storage prefix, and no other supplier will be able to read them.
  • Every release will record the agreement each of its files depends on, so the use of your data can be traced.
Talk to us about supplying data

For labs

For the data and research engineers who load the data, and the legal team that approves it.

  • A release will be a signed manifest, the data objects it lists, and a rights snapshot naming the agreement and classification behind every object.
  • A release will be copied into your own storage, and one command will verify every checksum and the manifest signature before it reports success.
  • Releases will be versioned and reproducible: the same version rebuilt will give the same manifest hash.
  • Acceptance tests will be agreed with you before processing at scale, and duplicate and public-overlap statistics will be stated in the dataset card.
Talk to us about a dataset

Company

GenSet Data is a new company building the platform described on this page. No supplier or lab uses it yet.

Coming soon. We are looking for the first supplier and the first AI lab to pilot it with. Write to us: team@gensetdata.com

Contact

General and partnership enquiries
team@gensetdata.com
Security issues, never in public
security@gensetdata.com