See if your data qualifies

Guides · Published

Best AI training data licensing companies (2026)

Compare six AI training data licensing providers in 2026, from marketplaces and rights-cleared image sets to business-record matching. See our criteria.

The best AI training data licensing provider in 2026 depends on what you have to license. Troveo and Depositphotos suit media and visual rights, Opendatabay and Versos help you compare license types and suppliers, and Smelted suits US companies that want to license internal business records such as support tickets, CRM history and project files.

We make Smelted, so it's on this list. We judged it on the same criteria as the others, using only what we could verify about it.

How we picked

Each provider is judged on five criteria:

  • Data type: what kind of material it takes, such as media, visual assets, or internal business records.
  • Cost to you as the data owner: whether you pay anything to take part.
  • Control: whether you keep ownership, choose what is included, and choose who receives it.
  • Rights clarity: whether the license is documented and explains what training is allowed.
  • Process and timing: how many steps there are and how long payment takes.

We reviewed the pages listed in Sources. Where a provider's page does not state a term, we say so rather than fill the gap. Credtent's neutral map of the market is a useful overview, but it is a guide rather than a seller, so it is not ranked here.

The list

1. Smelted

Smelted is a free matching service. It finds which AI data buyers fit a company, gives a ballpark range, and introduces the company to two or three matching buyers. The buyer pays Smelted's fee, and the company pays nothing.

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

The company keeps ownership. The buyer receives a license to a copy. The company chooses what is included and which buyers it meets, and names, contact details and IDs are removed before anything is shared. The value depends on headcount, years of records and how connected the data is, and a real figure comes only after a buyer reviews a sample.

What to watch for: payment usually comes a few weeks to two months after signing, so plan on one to three months from first call to payment. Smelted focuses on business records, so it is not the right place for audio, voice or video datasets. Buyers who take smaller companies pay less, so companies under 30 employees should talk to us directly first. To start, use the two-minute form at https://smelted.ai/get-started.

2. Troveo

Troveo runs a marketplace for licensed content. It covers video, audio, gaming, robotics and business data, and it pays rights holders under license agreements that name AI training as a permitted use.

Best for: rights holders with large libraries of video, audio or similar media who want a marketplace to handle many buyers.

Troveo reports that more than 7,000 licensors have signed agreements covering AI training, and that over $20 million has been paid out to rights holders so far. It also says 95 percent of licensors have signed exclusively, which matters if you want to keep the option of licensing elsewhere.

What to watch for: exclusivity is common on its platform, so read the terms before you sign. Its stated categories lean toward media, so check whether your records fit before you apply.

3. Depositphotos

Depositphotos sells rights-cleared stock imagery and datasets for AI use. Its blog describes AI data licensing as the legal authorization of datasets used to train machine learning and generative AI models.

Best for: teams that need visual data with clear rights for model training, and rights holders with image libraries.

What to watch for: its focus is visual content, so it is a weak fit for text records, code or engineering files. Its pages point to a custom plan rather than a published price, so ask for terms in writing.

4. Opendatabay

Opendatabay publishes a guide to licensed AI training data types. It compares commercial and general licenses for both buyers and sellers, and it covers how to choose one.

Best for: teams that want to understand license types before they sign anything, whether they are buying or selling.

What to watch for: the guide is educational rather than a single service with fixed terms. Read its license comparisons against the specific contract you are offered.

5. Versos

Versos publishes a list of direct suppliers of licensed AI training data. Its named suppliers include media and sports companies such as Getty Images and Shutterstock.

Best for: companies that want to see which established media owners already license content directly to AI buyers.

What to watch for: the list is built around media owners. It shows the market rather than a place to license your own internal records.

6. Ficstar

Ficstar's list of the best AI training data providers in 2026 includes HumanSignal, which offers Label Studio as open source plus managed services. The providers it covers handle image, video, text, audio and time-series data.

Best for: teams that need data labelling or annotation work rather than a license to existing records.

What to watch for: annotation is a service that prepares data, and licensing is a right to use data. Credtent's guide treats the two as separate, and you should check which one a provider sells before comparing prices.

Comparison

ProviderMain dataBest forCost to the data ownerWatch for
SmeltedInternal business recordsUS companies with 50+ employees and years of recordsFree to the company; buyer pays the feePayment 1 to 3 months after first call; no value until a sample is reviewed
TroveoVideo, audio, gaming, robotics, business dataRights holders with large media librariesNot stated on the pages reviewedExclusivity is common on its platform
DepositphotosRights-cleared images and datasetsVisual data with clear training rightsCustom plan; terms not publishedVisual focus only
OpendatabayGuide to licensed data typesLearning license types before signingNot stated on the pages reviewedEducational, not a single service
VersosDirectory of direct suppliersSeeing which media owners license directlyNot stated on the pages reviewedBuilt around media owners
FicstarLabelling and annotation servicesPreparing data for trainingNot stated on the pages reviewedAnnotation is not the same as licensing

FAQ

How do you sell your data to AI companies?

You either list it with a marketplace that matches rights holders to buyers, or you work with a matching service that introduces you to buyers. In both cases you should confirm that the license names AI training as a permitted use, which Troveo's guide describes as the core of a proper training license. Smelted anonymizes records before anything is shared, and you choose what is included and who you meet.

What is AI data licensing?

According to Depositphotos, AI data licensing in 2026 refers to the legal authorization of datasets used to train machine learning and generative AI models. The license defines what the buyer may do with the data and for how long.

How much money do companies make selling your data?

The amounts vary widely. According to Quartz, OpenAI's five-year licensing deal with News Corp is worth more than $250 million. Troveo reports that deals for mid-size publishers range from a few million dollars a year. Those figures involve large media owners, so they are not a guide to what a typical business's records are worth. For business records, the value depends on headcount, years of records and how connected the data is.

Licensing is generally handled through contracts that define permitted use, and Depositphotos covers the legal background, including fair use and copyright questions. This page is not legal advice, and no provider can promise legal or compliance approval for your specific data. Have counsel review any license before you sign.

Sources