See if your data qualifies

Guides · Published

Best AI training data licensing companies (2026)

The best AI training data licensing options in 2026, compared on what is licensed, cost to the seller, control and published evidence. Smelted is on the list.

For a company that wants to license its own internal business records to AI builders, Smelted is the best fit on this list. For media, video and audio catalogs, Troveo is the closer match. We make Smelted, so it's on this list; we judged it on the same criteria as the others, using only facts we can point to.

How we picked

We judged every entry on five criteria:

  1. What is licensed. Internal business records, media catalogs, or data services sold to buyers.
  2. Who is the seller. Whether a company with its own data can list it, or only rights holders with a catalog can.
  3. Cost to the seller. Whether the seller pays to take part, and how that is stated.
  4. Control. Whether the seller chooses what is shared and with whom.
  5. Evidence on the page. Whether the company publishes figures we can check.

We sorted companies by what each one licenses. A neutral guide to the sector makes the same point: lists of licensing companies often mix businesses that do different things. We left out companies whose main work is labeling or talent placement, and companies whose public pages gave us too little to judge on these criteria.

The list

1. Smelted

Smelted is a free matching service. It licenses anonymized internal business records, such as support tickets, CRM history, SOPs, project files, code, CAD and lab data, to AI data companies that train agents on them.

Best for: US companies with 50 to 500 employees and years of records in tools like Zendesk, Jira/GitHub, Salesforce/HubSpot, Bullhorn, AutoCAD/Revit or Benchling.

Strengths: The company pays nothing; the buyer pays Smelted's fee. The company keeps ownership and grants a license to a copy. It chooses what is included and which buyers it meets, and names, contact details and IDs are removed before anything is shared. Smelted introduces the company to 2 to 3 matching buyers and gives a ballpark range. Plan on one to three months from first call to payment.

What to watch for: The value depends on headcount, years of records and how connected the data is. A firm figure only comes after a buyer reviews a sample. Companies under 30 employees should expect that the buyers who take smaller companies pay less.

2. Troveo

Troveo runs a licensing marketplace for rights holders across video, audio, gaming, robotics and business data.

Best for: Rights holders with large catalogs of media or recorded content.

Strengths: Troveo reports more than 7,000 licensors who have signed agreements that explicitly cover AI training, with more than $20 million paid out to rights holders so far. Troveo also states that 95 percent of licensors signed exclusively, which gives a clear picture of how its contracts are set up.

What to watch for: Troveo's material centers on media and rights holders. Its page does not state what a seller pays or how much control a licensor keeps over each buyer, so check those terms before signing. Reported deals range widely; Troveo cites a News Corp and OpenAI deal at $250 million over five years, which is a large media agreement, not a benchmark for internal records.

3. Opendatabay

Opendatabay publishes a guide to AI training data licenses, aimed at buyers and sellers.

Best for: Companies that want to understand license types before they list anything.

Strengths: Its guide compares commercial and general licenses for both sides of a deal, which helps a seller see what they are granting before they agree.

What to watch for: The public guide we reviewed describes licenses; it does not describe a sign-up process, fees or the controls a seller keeps. Read the terms of any specific listing.

4. Ficstar

Ficstar is a data collection and AI training data provider that has run web data work for more than 20 years.

Best for: Buyers who need data collected or labeled at scale, not sellers looking for a place to license their own records.

Strengths: Ficstar states that it serves more than 200 enterprise clients and processes over 1 billion product prices every month. Its guide also explains how to test a provider, which is useful when you compare services.

What to watch for: Ficstar sells data services to buyers. It is not a route for a company to license its internal records, so it fits a different need.

5. Depositphotos

Depositphotos is a stock media company that publishes on rights-cleared AI data licensing.

Best for: Image and media owners who want to understand rights-cleared licensing.

Strengths: Its guide explains how rights-cleared datasets reduce legal risk in AI projects, a topic the other entries address less directly.

What to watch for: The material is about image and media content. It does not cover internal business records, and its page does not state terms for sellers.

Comparison

EntryWhat is licensedCost to the sellerControl over sharingEvidence on the page
SmeltedAnonymized internal business recordsFree; buyer pays the feeSeller chooses data and buyers; names and IDs removedProcess described; value set after a sample review
TroveoMedia, audio, gaming, robotics, business dataNot stated on the pageLicensors sign per agreement; 95% exclusive7,000+ licensors; over $20 million paid out
OpendatabayLicensed datasets (guide)Not stated on the pageNot stated on the pageComparison of license types
FicstarServices: collection and labeling for buyersNot a seller routeNot applicable20+ years; 200+ enterprise clients
DepositphotosRights-cleared image and mediaNot stated on the pageNot stated on the pageGuide on rights-cleared licensing

How big is the AI training data market?

The market is growing and North America holds the largest share. According to Grand View Research as cited by Troveo, the global AI training dataset market is projected to reach $16.3 billion by 2033, growing at 22.6 percent a year from 2026. Image and video data held a 41.9 percent revenue share in 2025, and North America had 35.1 percent of global revenue.

The demand for non-public data is rising. Epoch AI projects that the stock of high-quality public text data could be effectively used up for training between 2026 and 2032, which is one reason labs look to licensed data. MarketsandMarkets puts the AI training dataset market at about $2.82 billion in 2024, projected to reach $9.58 billion by 2029.

FAQ

How do you sell your data to AI companies?

You list anonymized data with a matching service or marketplace, and you choose what is included. Smelted works this way for business records: a company names no one, gives no contact details, and keeps ownership while it grants a license to a copy. The company completes a 2-minute form, Smelted finds matching buyers, and the company decides which buyers it meets.

How much money do companies get for selling your data?

We do not give a dollar figure in advance. The value depends on headcount, years of records and how connected the data is. Once a buyer reviews a sample, the figure becomes firm. Reported media deals are much larger; Troveo cites a $250 million, five-year News Corp and OpenAI deal, which is not a benchmark for a typical company.

How much is 100 GB of data worth?

Size alone does not set the value. What matters more is what the records contain and how connected they are, along with headcount and years of history. A buyer reviews a sample before any real number is set.

Sources