Who buys company data for AI: buyers, aggregators and brokers

What are the four kinds of program?

Programs that pay for company data describe themselves in different words: partnerships, licensing, data assets. We sort them by what they do with the data and who pays whom, because that is what changes your position. These are the four kinds we use in the buyer comparison.

Kind	What it does	Who pays the company
Buyer	Licenses data for its own AI training and evaluation work	The buyer
Aggregator	Pays the company, then licenses the data on to AI labs	The aggregator
Broker	Introduces companies to buyers for a fee	The buyer, directly
Closure marketplace	Finds licensees or purchasers for the assets of closing companies, data included	Proceeds pass through the marketplace

Buyers sign with you, pay you and use the data for their own work. Aggregators also pay you, but their business is licensing your data on to AI labs, and their structures differ: Endless Data agrees a fixed sum, Polyshares pays a single agreed amount and Turing pays a share of each license. Brokers never pay you; the buyer they introduce does. SellData.ai works the same way: we introduce companies to buyers, and the buyer pays you. Closure marketplaces such as SimpleClosure AssetHub focus on closing companies and can arrange a license or an outright sale.

Which programs pay for company data today?

These are the programs we track as of November 9, 2026. Each line is ours and stays within what the program publishes; program pages have every field with sources.

Program	Kind	In one line
micro1	Aggregator	Company Data Partnerships program for established companies, licensing to labs and enterprises in the US and allied countries, with a separate track for companies that are no longer operating
Handshake AI	Buyer	Data Partnerships program for AI training and evaluation, separate from its Fellowship expert program
Mercor	Aggregator	Connects to company tools, de-identifies the data and licenses it to frontier AI labs
Scale AI	Aggregator	De-identifies operational records inside the company’s own environment and turns them into training data for AI companies, as an ongoing partnership
Polyshares	Aggregator	Licenses the operational record of operating companies, de-identifies it through a PII partner and offers it to AI buyers through its own portal
Endless Data	Aggregator	Pays operating companies a fixed sum and licenses the records to AI labs it names as micro1, Mercor and Handshake
Mode	Aggregator	Buys an agreed copy of company operating records for businesses and nonprofits, de-identifies it and resells it for AI research and training
Nyne AI	Aggregator	Licenses company operational data, de-identifies it and delivers it to frontier AI labs under confidentiality, alongside an expert network
Turing (Project Lazarus)	Aggregator	Prepares a company record from exports, licenses it to buyers and pays the company a share of each license
SimpleClosure AssetHub	Closure marketplace	Licenses source code and workspace data, or arranges an outright sale, for closing and operating companies

Where does your data end up with each kind?

The kind decides who holds a copy after the deal. That matters more than the headline number, because once a model has been trained on a copy, deleting the copy does not undo the training.

Kind	You sign with	Who ends up with a copy
Buyer	The buyer	The buyer, plus any third parties its agreement allows
Aggregator	The aggregator	The aggregator and its customers, which can be several AI labs
Broker	The buyer the broker introduces	The buyer you choose; the broker holds no data
Closure marketplace	Set in each deal	The licensee or purchaser you approve

Buyer. Handshake AI says use is restricted by contract to AI model training and evaluation (source). Being a buyer does not by itself mean the data stays with the buyer, so read the sharing and resale clauses.

Aggregator. Licensing on to others is the business model. micro1 says data is licensed to labs and enterprises in the US and allied countries, that personal and identifying information is removed or transformed before sharing with third parties, and that its contract commits it not to resell until the data is anonymized (source). Mercor says no AI lab buyer or third-party vendor sees your raw data; the de-identified version is what reaches labs (source). Turing pays you from each license, so a revenue share rewards more licensees, which also means more copies (source). Polyshares says its licenses are exclusive by default unless agreed otherwise (source); Endless Data says its licenses are non-exclusive (source). Endless Data names micro1, Mercor and Handshake as the labs it licenses to, so the same records can pass through more than one program on this list. Its buyer terms also let its customers keep models trained during the term after the data is deleted (source).

Broker. A broker sits outside the data flow. Your data ends up with whichever buyer you sign with, under that buyer’s terms.

Closure marketplace. SimpleClosure says assets reach buyers only with the owner’s approval, and that it handles matching, documentation and distribution of proceeds (source). It offers a license, where you keep the assets, or an outright sale, where you do not.

How can you tell them apart?

Labels on websites are not a reliable guide. A quick test: if the party pays you itself, it is a buyer or an aggregator, and the next question is whether it licenses your data on to others. If it never touches the money, it is a broker or advisor, and the next question is who pays it.

Questions worth asking any program or intermediary:

Question	Why it matters
Who pays us, and on what trigger?	Payment on signing, delivery and acceptance are different promises
Who else will receive a copy, now or later?	Tells you whether you are dealing with a buyer or an aggregator
Do you take any part of our price?	Success fees and buyer-side fees change what you net
Are you paid by the programs you recommend?	Any advisor should answer this in writing

Where should you start?

Start with fit, not with the kind. Our value check compares your headcount, history, tools and status with each program’s published criteria, and the buyer comparison shows them all in one place. For two programs side by side, see Mercor vs micro1; if the company is closing, see licensing data from a closed startup. For the full deal process, read how licensing company data to AI works.