Short answer
Four kinds of buyers pay for AI training data: AI labs that build large models, specialist data suppliers that license business data and supply the labs, established technology companies building their own models, and marketplaces that resell. The route depends on the buyer, dataset and program. Check the counterparty and permitted end uses.
If your company has data that AI developers want, the first practical question is who would actually pay for it. The direct buyer may be a lab or a specialist supplier. This guide maps the buyers, explains where each sits in the supply chain and lists the questions to ask before dealing with any of them.
Key takeaways
- There are four broad kinds of buyers: AI labs, data suppliers, technology companies building their own models, and resale marketplaces.
- AI labs sign large deals directly with big content owners. Other programs source records through data suppliers.
- A company may deal with data suppliers: firms that license business data, prepare it and supply the labs.
- You should always know who the buyer is, what it will do with the data and how the data will be de-identified.
The four kinds of buyers
| Buyer | What it does | Buys from midsize companies directly? |
|---|---|---|
| AI labs | Build and train large AI models | Depends on the program and dataset |
| Data suppliers | License business data, de-identify and package it, supply AI labs | Yes, this is their business |
| Technology companies | Build models for their own products | Sometimes |
| Marketplaces and wind-down firms | List data from many sellers, often companies that are closing | Yes, mostly shutdowns |
AI labs
Labs are one possible end user of licensed training data. Public content-licensing deals include agreements with Google and OpenAI. These examples do not identify the buyers or end users in any DataRelai program.
Publicly reported direct deals include large content owners. Reddit's agreement to license its content to Google was reported at about $60 million a year. News Corp's multi-year agreement with OpenAI was reported to be potentially worth more than $250 million over five years, though the companies did not disclose terms. Those are corporate deals negotiated by large teams. See publicly reported AI data licensing deals.
For operational business data, a supplier may sit between the company and the end user.
Data suppliers
Data suppliers are one possible direct counterparty. They find operating businesses with the right archives, license the records, de-identify and package them, and supply the result to AI labs for training models and AI agents.
Suppliers may license operational records as principal and supply them onward. Ask who the direct buyer is, whether onward use is allowed and which end users and uses are covered by the contract.
Suppliers differ in important ways: what they pay, whether their license is exclusive, how they de-identify data and how hard they push on deadlines. That is why comparing more than one matters.
Technology companies building their own models
Not every buyer is supplying someone else. Technology companies building models are another possible category to ask about. These buyers may want narrower data that matches their own products, such as records from a specific industry.
Marketplaces and wind-down firms
A separate channel serves companies that are shutting down. Firms that help startups close have begun listing the code and workspace data of defunct companies for sale. That does not make an operating company's records equivalent to a wind-down asset; ongoing contractual and employee obligations still matter.
This channel is built for shutdowns. An operating company has more to protect, including customer relationships, employee trust and future acquisition value, and usually needs a more careful process.
How to vet an AI data buyer
Ask every buyer the same questions and get the answers in writing.
- Who are you, and who is the end user of the data? You should always see who the buyer is and what it is offering.
- What exactly will the data be used for? Training only, or wider rights such as resale?
- How is the data de-identified? Who does it, to what standard, and who verifies it? See de-identification for AI training data.
- Is the license exclusive? For how long? See exclusive vs. non-exclusive data licenses.
- When is payment made? On signing, on delivery or on acceptance?
- What happens to raw data during review? Records should move only under a signed license.
Warning signs include pressure to share sample data before any agreement, offers that expire within a day or two, vague answers about the end user and no written description of de-identification.
Where a seller-side consultant fits
Buyers do this every week. Most companies will do it once. A seller-side consultant narrows that gap by knowing which buyers want what, and by running the comparison for you.
DataRelai works for the seller as your consultant, never the buyer. We qualify your data, find buyers and bring you their offers. We only work with buyers who anonymize and de-identify data, and you approve every deal before anything moves. The fee is a 10% success fee, only when a deal closes and you've been paid. We may also receive a referral fee from a buyer. For how this role differs from a broker's, see data broker vs. data licensing consultant.
Frequently asked questions
Do OpenAI or Google buy data directly from small businesses?
It depends on the program and dataset. Public lab deals include large content owners; a company may also deal with a specialist supplier. Verify the counterparty and permitted uses in the license.
How do I find AI data buyers?
Several suppliers advertise programs publicly, and some approach companies directly. A seller-side consultant can match your data to buyers looking for it and bring you their offers.
Will the buyer tell me which AI lab gets my data?
Not always. Supply contracts between data suppliers and AI labs are rarely disclosed. You should still know who your direct buyer is and what uses your license permits.
Are AI data buyers legitimate?
Many are established businesses, but the market is young and practices vary. Vet each buyer, compare offers and have your own attorney review the agreement before you sign.
Sources
- Engadget, "Reddit is licensing its content to Google to help train its AI models" (February 2024)
- Associated Press via WSLS, "OpenAI to start using news content from News Corp. as part of a multiyear deal" (May 23, 2024)
This guide is general information, not legal, tax or financial advice. Whether a particular dataset can be licensed depends on what it contains and on obligations your company already has.