Who buys company data?

Frontier labs training general purpose models sit at one end of the chain. They need operating material at scale and hold the strictest standard for provenance in the market. Applied AI companies building for a single industry sit alongside them, and for a specialized record they often want it more badly. Neither group is set up to negotiate with a fifty person business in a specialized trade, so the buying happens through data labs instead. A data lab sources the material, verifies that the seller has the right to license it, clears and prepares it, and supplies the labs. Polyshares is a data lab, and our focus is enterprise operating records rather than anything produced to order.

For the company selling, this means no introduction, no listing, no auction and no queue. You sign one license, with us, and we pay you. What happens downstream is our problem and our risk, and if we misjudge what the record is worth, that is our loss rather than a renegotiation with you. This is the practical difference between a buyer and an agent, since an agent is paid out of your proceeds and carries none of the exposure.

Any buyer checks provenance first, before price is even discussed, because nobody with regulatory exposure signs without a clean chain showing the seller had the right to grant the license. It is the most common reason a promising record goes nowhere. After that comes whether personal information can be removed properly, and how much of the record is genuinely unpublished rather than already scraped from the open web.

Scope and permitted use are written into the license, and Polyshares controls where the material goes and does not disclose which labs receive it. A company that lists on an open marketplace gives up that control entirely. A bounded, non-exclusive license also leaves the company free to license the same record again as it grows, to Polyshares or elsewhere.

Updated September 2026.