Which AI training data companies buy company data?

AI training data companies that buy records from businesses fall into three groups: data partnership programs (micro1, an AI data company, and Polyshares), sector specialists (Protege, Troveo) and data marketplaces (AWS Data Exchange and others). The two partnership programs buy general operational records. micro1 asks for 30 or more full-time employees, and Polyshares states no headcount minimum.

micro1 pays us a referral fee only after a company we refer has its data transferred and approved.

All facts below are as of November 2026. Pay figures are the buyers’ own claims, so treat them as ranges.

How company data reaches an AI lab

AI labs need data they can’t scrape: how real companies actually run. Data companies get it by paying operating businesses to license their operational records, then de-identifying and packaging the records for the labs.

That puts three parties in the chain. Your company holds the records. A data company licenses them from you and prepares them. An AI lab receives the prepared data from the data company. The business sells a license and keeps its data.

Most questions about a buyer come down to what the data company agrees with you: who may use the records, for how long, who removes personal data and when you are paid.

The three kinds of buyer

A company with years of SOPs, tickets and CRM history is the kind of seller the partnership programs describe, because they take general operational records and do the packaging themselves.

micro1 and Polyshares compared

A run rate projects recent activity forward. It is not booked revenue, and the figure comes from an unnamed source.

An exclusive, perpetual license generally means you cannot license the same records to anyone else, so the price has to cover the alternatives you give up. Ask every buyer whether its license is exclusive. micro1’s page says the data is licensed exclusively to labs and enterprises in the U.S. and allied countries, which limits who receives it. It doesn’t say whether your own rights are exclusive, so ask before you sign. Our review of its program lists the questions to put to it.

Treat both pay ranges as the buyers’ own claims. Both buyers say the depth of your history matters, and micro1 adds quality and uniqueness. Our guide to data valuation covers the factors in detail.

Which buyer fits which company

These are starting points. Each buyer decides case by case and can change its rules, so confirm on its website before you invest time in preparing data.

What the published figures can and can’t tell you

Every figure on this page is a buyer’s own statement, as of November 2026. None is an audited result, and none says how many applicants were turned down. Neither buyer publishes an average or median deal that we have found. The Polyshares deals that are published ($110K to $450K) sit in the lower part of the $100K to $2M and above range its site gives for typical deals, so don’t assume the top of a range is typical.

The figures also leave out timing. Approval can take months after a company applies, which is our own estimate. Use the numbers to decide whether a conversation is worth having, then ask the buyer for terms that fit your own records.

Sector specialists

Protege focuses on health and media data, and Troveo on video, audio and games. A company whose records are mainly clinical notes, media archives or footage may find a specialist a better match than a general program. Check each buyer’s website for the record types it accepts and the terms it publishes: Protege and Troveo.

Data marketplaces

A marketplace connects data providers with buyers. The platforms in our research are AWS Data Exchange, Snowflake Marketplace, Databricks Marketplace, Datarade, Nomad Data, Opendatabay and Defined.ai.

To use one, you prepare a dataset as a product, set a price and license terms, and let buyers find it. Each platform publishes its own fees, so read the fee schedule before you list. Our guide to data monetization compares marketplaces with direct licensing.

Other AI training data companies that sell to labs

Scale, Surge, Mercor, Handshake, Turing and Snorkel also work in AI training data. They sell data and services to AI labs. In our research, none of them runs a program that buys operational records from companies, so they are not on the buyer list above.

How to compare buyers

1. Size rule. micro1 asks for 30 or more full-time employees. Polyshares’ site says any company with real operating history, in any sector, qualifies and states no headcount minimum.
2. Record types. Ask each buyer which types it accepts and which it excludes. Buyers generally exclude data you hold for clients: if it sits under a client contract or NDA, it usually isn’t yours to license.
3. Exclusivity and duration. An exclusive, perpetual license means you give up the right to license the same records elsewhere, so the price has to reflect that. Our guide to data licensing explains the options.
4. Permitted use. Ask what the buyer and the labs it supplies may do with your records, and whether you can see who receives them.
5. Who removes personal data. Polyshares uses a specialist partner. micro1 says companies need not anonymize before transfer and that it has processes for de-identification. Get the method, the review step and the liability in writing.
6. Payment. Ask for the amount, the schedule and the conditions. The published figures are ranges, and marketplaces publish their own fees.
7. Ending the deal. Ask what must be deleted or returned if the license ends, and who confirms it.
8. What you keep. A license lets the buyer use a copy of your records under agreed terms. You keep your records, though an exclusive license limits what else you can do with them.

What to have ready before you contact a buyer

A first conversation goes faster when you can answer a few plain questions:

• How many people work at the company, and in which countries does it operate?
• Which record types do you hold, in what language, and how many years does each go back?
• Roughly how much does each type hold (documents, tickets, accounts), and where is it stored?
• Do any of the records belong to clients or sit under an NDA?
• Who has the authority to sign a license for the company?

Nothing has to be cleaned or shared at this stage. micro1’s FAQ says companies need not anonymize data before transferring it, and any buyer you approach should be willing to explain its own sequence.

How we list buyers

The facts about each buyer come from its published material and our own research, and they carry an “as of November 2026” date. Weigh how prominently we cover micro1 against the disclosure at the top of this page. We have no referral arrangement with Polyshares, Protege, Troveo, Defined.ai or the marketplaces. Our disclosure page explains how we make money.

Next steps

For the full process, from ownership check to payment, see how to sell your company’s data to AI companies. If your company has 30 or more employees, the eligibility check below asks four questions.