Is This AI Data Buyer Legit? A 10-Point Checklist

Is this AI data buyer legit? The quick test

Since 2025 AI labs have paid real money for company records, and that has drawn in new buyers, resellers and sites that look like buyers. Most are honest; some are new companies with no money behind them, and a few are scams. You cannot tell them apart from a landing page.

Five questions settle most cases. Who signs the license? Who ends up with the data? What is the price, and what must happen before it is paid? Who removes personal data, and when? Does anyone want money or passwords from you? If a buyer answers all five plainly and in writing, the rest of this checklist is detail. If it dodges any of them, treat that as your answer.

SimpleClosure, which runs startup wind-downs, describes speaking with "a would-be buyer whose company had launched only weeks earlier" (SimpleClosure). New is not the same as fraudulent, but a seven-figure promise from a company with no history needs proof.

The 10-point checklist

Work through these in order. The first three take minutes and rule out most bad actors.

• 1. Verify the entity. Look the company up in its state registry; in Delaware, the Division of Corporations search returns active and inactive entities. Check founders, funding news and how long the domain has existed.
• 2. Ask who the end buyer is. A vendor may not name the lab, but it should say what kind of company receives the data and in which countries. micro1, for example, states that data is licensed only to labs and enterprises in the US and allied countries.
• 3. Ask for completed deals. A real buyer can describe deals it has closed, even anonymously. Polyshares publishes four, with prices and license terms.
• 4. Get the price and payment trigger in writing. The contract should state the amount, what must happen before payment and the date it arrives. An estimate or an ad is not an offer.
• 5. Check exclusivity and term. Exclusivity is a normal term: Labelbox Prism licenses exclusively by default, and Polyshares' published deals were perpetual exclusive licenses. The risk is exclusivity with no end date and no premium over a non-exclusive price, so ask what the non-exclusive price would be. See exclusive vs non-exclusive licenses.
• 6. Ask who de-identifies the data, where and when. The safest answer is before anything leaves your systems, with a sample you approve. Ask what categories are excluded: HR, payroll, privileged legal material and credentials should be.
• 7. Control the access method. A scoped export, or a read-only connection limited to the agreed tools that you can revoke, set up under a signed contract, is normal. Admin passwords, unscoped credentials or any access before the license is signed are not.
• 8. Read the acceptance, holdback and clawback terms. If payment depends on the buyer accepting the data, the acceptance test and its deadline must be written down, along with what happens to rejected data and whether any money can be taken back.
• 9. Refuse upfront fees. A buyer pays you. Any request for a valuation fee, a listing fee or a "processing deposit" before a deal is a red flag.
• 10. Ignore pressure. Offers that expire in days, demands for a full export before an offer, or refusal to let your lawyer read the contract are reasons to stop.

Red flags in a data licensing offer, and what a sound buyer does instead

Each "sound buyer" example below is taken from a program's own page, read on 10 November 2026. Naming a program here is not an endorsement; it shows that the practice is normal.

Is it safe to sell company data?

It can be, if three things are true: personal and confidential data is removed before the records leave you, you control what is shared and can stop sharing, and the license limits what the buyer may do with the data. The risks are leaks of customer or employee information, breaches of contracts with your own customers, and losing control of data once it is resold.

Two practical points from people who handle these sales. Hidden secrets: SimpleClosure warns about "Live credentials such as API keys and access tokens, which can be buried in code, tickets, and attachments." Order of work: "If the buyer receives the raw files first and promises to clean them up later, you have already given that buyer access to the material you meant to remove" (SimpleClosure). And the end use is often unclear: MIT Technology Review found that "Founders have little to no visibility into who is buying the data and what they plan to do with it" (MIT Technology Review).

Ask for a re-identification test before delivery. AfterQuery says every dataset gets one, "because no process removes all risk on its own".

Source: SimpleClosure: how real are those million-dollar data offers? (Oct 2026)

Is selling company data legal?

Yes, in most cases, when the company owns the records and has the right to license them. The legal questions are about the contents: personal data about customers and employees, information covered by customer contracts or NDAs, licensed third-party material and open-source code with its own terms, and regulated data such as health records.

Most AI data deals are licenses, not sales: you keep ownership and grant defined rights for a defined use and term, paid either once or as a share of each sale. Read is it legal to sell company data for AI training, is selling company data to AI legit and the data licensing agreement guide before you sign. A lawyer should read any license over a few thousand dollars.

How to check a data licensing company in an hour

This is the order Origin uses when an owner asks us to look at a buyer.

• Registry: search the legal name in its state registry and confirm it is active and matches the name on the contract.
• People: find the founders and check what they have built before and whether they are reachable.
• Money: look for a named funding round reported by a news outlet. Mercor, micro1, Scale AI, AfterQuery and Turing all have reported rounds.
• Track record: ask for completed deals and, if possible, a reference from a company that was paid.
• Contact channel: confirm that emails come from the company's own domain. Appen, for example, says "All official emails will be sent from @appen.com email addresses."
• Program page: compare what the buyer told you with what its own page publishes. Any difference in price, floor or timing is a question to ask.
• Paper: get the license in draft and check it against our due-diligence guide.

AI data buyer scams: what they look like

Most problems are not outright fraud but inflated promises. SimpleClosure notes that "An ad offering up to $5 million doesn't tell you what this buyer will pay for your assets" (SimpleClosure). Nyne gives sellers the same advice: "Separate contracted cash from advertised payout ranges, estimates and possible future revenue" (Nyne).

Outright scams follow patterns known from job and payment fraud. Appen warns that it is "seeing an increase in scams targeting job seekers and businesses" and states "We will never ask you to send money to apply or start a job with us" (Appen). Watch for impersonation of a known buyer from a look-alike domain, requests for fees or deposits, checks sent in advance, and requests for system logins. If an approach uses a known company's name, contact that company through its own website to confirm it.

Intermediaries are a third case. Some present themselves as buyers but pass your details to a program that may pay them. That is legal, but you are entitled to know it: ask any intermediary, Origin included, who the buyer is, what the terms are and how it is paid. Our AI data companies directory lists the programs that buy directly.

Source: Appen: scam awareness

Get a second opinion before you sign

You do not need to run this checklist alone. Origin reviews offers for owners: we check who the buyer is and who ends up with the data, read the license terms that move the price, and compare the number with published deals for similar records. If the offer is sound, we say so; if a better buyer exists for your records, we tell you that too. Know one thing first: a buyer may pay Origin when a deal closes, as the note below explains. Compare programs first in Mercor vs micro1.

Not sure whether a buyer or an offer is sound? Origin checks who the buyer is, what the license allows and how the price compares with published deals before anything is signed. Get a free estimate and a second opinion.

Origin Data Partners works with companies that own data and helps get it bought. We are independent of the services on this page, and a buyer may pay us when a deal closes (how Origin is paid). See what your data is worth.