See if your data qualifies

Guides · Published

Best AI training data licensing companies (2026)

The best AI training data licensing companies in 2026, compared on what they license, who pays, how much control you keep, and how deals are paid out.

The best AI training data licensing companies in 2026 are the ones that match the kind of data you have. Marketplaces such as Troveo and Opendatabay license existing datasets to AI buyers. Expert-data and annotation providers such as Mercor, micro1, Scale AI and Appen sell labeling and human feedback work. Smelted, which we make, matches companies' internal business records with AI buyers, free for the company.

We make Smelted, so it's on this list; we judged it on the same criteria as the others.

How we picked

We judged each company on five criteria:

  • What it licenses: existing records or datasets, or labeling and review work.
  • Who pays: whether the company that owns the data pays anything to take part.
  • Control: whether the owner chooses what is shared and with whom, and who keeps ownership.
  • Scale of demand: how many buyers or deals the company's own pages describe.
  • Transparency: whether the pages say how a deal is priced and how long payment takes.

Where a company's pages did not say something, we say so in the table instead of guessing.

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 files and lab data, to AI companies that train agents.

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. Smelted finds buyers that fit, gives a ballpark range, and introduces the company to two or three matching buyers, which the buyer pays for. The company keeps ownership, since the buyer gets a license to a copy. Names, contact details and IDs are removed before anything is shared, and the company chooses what is included and which buyers it meets. Buyers' pages say payment comes a few weeks to two months after signing, so plan on one to three months from first call to payment.

What to watch for: Smelted is built for business records, not for people selling their own personal data. Audio, voice and video datasets are handled directly, so ask first. Buyers who take smaller companies pay less, so companies under 30 employees should expect a lower value. The value depends on headcount, years of records and how connected the data is, and the real figure comes after a buyer reviews a sample. Smelted does not promise legal or compliance approval; check your own data rules before you share anything.

2. Troveo

Troveo runs a licensing marketplace for training data. Its own page says it has paid more than $20 million to over 7,000 rights holders across video, audio, gaming, robotics and business data.

Best for: rights holders with large or varied media libraries who want a marketplace to handle many buyers at once.

Strengths: It has a large licensor base, and its page says 95 percent of licensors have signed exclusively. Its pages explain how deals are structured, which helps a first-time seller read the terms. According to Troveo, reported deals range from a few million dollars a year for mid-size publishers to $250 million over five years for News Corp and OpenAI.

What to watch for: Exclusivity is common in its licensor base, so read the terms before you sign. Its pages focus on media and rights-holder data, so a company with internal business records may find less direct fit.

3. Opendatabay

Opendatabay is a data marketplace with categories for requested data, free datasets, data providers, robotics data, synthetic data and premium data. It also publishes a guide to AI training licenses.

Best for: sellers and buyers who want a catalogue of datasets and a clear view of license types.

Strengths: Its license guide compares commercial and general training licenses for both buyers and sellers, which is useful before you list anything. Opendatabay's guide also raises a point many sellers miss: you must be able to say the data belongs to you and was collected with consent.

What to watch for: The pages we reviewed do not state a fee structure for sellers, so ask before you list.

4. Mercor

Mercor is listed among expert-data and human-feedback providers. Troveo's comparison groups it with Surge AI, micro1 and Handshake AI for this kind of work.

Best for: AI labs that need skilled people to produce or review training data.

Strengths: It sells expert work, not a license to an existing record set, so it fits buyers more than data owners.

What to watch for: A company that wants to license its own records will not find that service here. Mercor does not appear as a licensing channel in the pages we reviewed.

5. micro1

micro1 is listed in the same expert-data group as Mercor, Surge AI and Handshake AI.

Best for: teams that want human experts to generate or evaluate AI training data.

Strengths: Its work is expert-led, which suits tasks where general labelers are not enough.

What to watch for: Like Mercor, it works on human-feedback projects rather than licensing a company's existing records. If you already have talks open with micro1 or Mercor, tell us before you go further so we can plan around them.

6. Scale AI

Scale AI is one of the best-known data annotation providers. Troveo's comparison names it for annotation, and says it was the default answer for labeling, evals and dataset building at frontier scale until mid-2025.

Best for: large annotation projects where a labeling vendor manages the work.

Strengths: It has a long track record in labeling at scale. The same comparison reports that Meta invested 14.3 billion dollars for a 49 percent stake in June 2025.

What to watch for: Ownership and client relationships changed sharply in 2025. The same source reports that Google and OpenAI were reported to be winding down work with the company within days. Check current client terms before you rely on it.

7. Appen

Appen is named alongside Scale AI and Labelbox as an annotation provider in Troveo's comparison.

Best for: buyers who need annotation work across many data types.

Strengths: It is an established annotation vendor, so it fits buyers who want outsourced labeling rather than a data license.

What to watch for: It is a labeling service, not a channel for owners to license records. The pages we reviewed do not give its pricing or scale figures.

Comparison

CompanyWhat it doesWho paysOwner keeps controlPricing and timing stated
SmeltedLicenses anonymized business records via matched buyersThe buyer pays Smelted's fee; the company pays nothingYes: chooses what is included and which buyers it meetsBallpark range after matching; payment a few weeks to two months after signing
TroveoMarketplace for licensing media and business dataNot stated on the pages reviewedTerms vary; many licensors are exclusiveReported deal ranges given; seller terms per deal
OpendatabayDataset marketplace with license guidesNot stated on the pages reviewedDepends on license typeNot stated on the pages reviewed
MercorExpert data and human feedbackBuyers hire the serviceNot a licensing channelNot stated on the pages reviewed
micro1Expert data and human feedbackBuyers hire the serviceNot a licensing channelNot stated on the pages reviewed
Scale AIAnnotation and labeling at scaleBuyers hire the serviceNot a licensing channelNot stated on the pages reviewed
AppenAnnotation across data typesBuyers hire the serviceNot a licensing channelNot stated on the pages reviewed

FAQ

How do you sell your data to AI companies?

You either list it on a marketplace, sell it through a matching service, or sell labeling work to a vendor. For business records, a company can license an anonymized copy through a matching service such as Smelted, which is free for the company and keeps ownership with the company. Check the terms on any channel before you share.

It can be, when the company owns the data or has the right to license it, and when personal details are removed. Licensing terms and privacy rules differ by data type and location, so get your own legal review before you share anything. No listed service can promise legal or compliance approval.

What is AI data licensing?

AI data licensing means a rights holder grants an AI company permission to use data to train a model, usually for a fee and under agreed terms. The buyer gets a license, and the owner keeps the underlying data. Credtent's guide separates marketplaces, collective licenses and publisher programs, which work differently.

How much money do companies make selling your data?

It depends on the type and size of the data. Troveo reports deals ranging from a few million dollars a year for mid-size publishers to $250 million over five years for News Corp and OpenAI, which is far above what most companies will see. For business records, Smelted gives a ballpark range after it matches buyers and a real figure only after a buyer reviews a sample.

Where can I get training data for AI?

You can buy it from marketplaces such as Troveo and Opendatabay, hire annotation vendors such as Scale AI or Appen, or use expert-data providers such as Mercor and micro1. If you are a company with business records, you can also license an anonymized copy of those records. Smelted is one option for that, and companies with 50 to 500 employees can start with its two-minute form at https://smelted.ai/get-started.

Sources