Updated 15 Oct 2026
Who sells AI training data, and where each gets it.
150 companies sell training data, RL environments, evaluations or expert work to the companies that build AI. Each line here is what the company says on its own site, with the sentence it rests on. 17 say they build from data licensed from the people and companies that hold it, and 29 say they pay for data from outside.
150 vendors, A to Z. Not ranked.
- How it is made.
- Every company was read on its own site, and a script checked each quote word for word against the page; a claim whose quote did not match is left out. Funding, revenue and acquisitions come from the company records, never from the vendor's own copy.
- Not a ranking.
- No vendor pays to be here or chooses what its row says, and none is our client. What a vendor says it sells is its own claim: its company page shows what others have confirmed.
- Why a holder reads it.
- A vendor that builds from licensed data has to buy it from someone. The ones that say so, with the quote, are buyers beyond the 37 with a published programme on the buyer board.
Most vendors make their data to order. A few start from records someone already has.
Most vendors here pay experts and contractors to write tasks, demonstrations and judgments, or generate them. A smaller group starts from records that already exist: a company's tickets and code, a studio's footage, a hospital's de-identified files.
That second group needs holders. Filter by “Licensed from its owners” or “Says it pays for data” to see who they are, in their own words.
Cite this
AIData.Broker, “Who sells AI training data, and where each gets it”, reviewed 15 Oct 2026. https://aidata.markets/vendors
Free to reuse under CC BY 4.0 if you credit AIData.Broker and link to this page. Each row names the publication that stated it, and words quoted from a source keep that source's own terms.