Reported Payouts to Data Sellers: Advertised vs Confirmed

Most payout figures for data sold to AI companies are published by the company that pays them, and the amounts a seller, a filing or a court confirms are few and rarely match a program's headline. Handshake states USD 250 million "paid out to partners" (re-read 2026-11-10), while one named founder reports roughly USD 25,000 net for a codebase (MIT Technology Review, 2026-10-13). Both are dated, and neither confirms the other.

Payout reports are the least verified numbers in AI training data licensing. This page sorts them by who states them, lists the paid-out totals that buyers and intermediaries publish, and sets them against amounts that sellers, filings and courts confirm, as of November 10, 2026.

DataPacts Editorial Team checked every figure on this page against its primary source on November 9, 2026 (Handshake and Troveo re-read November 10); the page on how we source, date and check every figure sets out the rules.

Four Kinds of Payout Figure, From Advertised to Confirmed

A payout figure is only as strong as the party that states it: an advertised range is a sales message, a claimed paid-out total is the payer's own count, a contract value or bid is a promise on the record, and a confirmed amount is one the recipient, a filing or a court states was paid. The table gives one example of each, as of 2026-11-10.

Only the last level shows money received. The first three can all hold while no recipient has been paid what they suggest, so each figure below keeps its level, stating party and date. The rule is set out under how we source, date and check every figure.

Paid-Out Totals That Buyers and Intermediaries Publish

Eight companies publish a running total of what they say they have paid, and every one of them is the payer's own figure, checked November 9, 2026 (Handshake and Troveo re-read November 10), with no recipient named. The table lists them alphabetically, and none of the totals is summed, averaged or ranked. Troveo's USD 850,000 payout card (re-read 2026-11-10) is labeled "ILLUSTRATIVE EXAMPLE" and is not a payment, and micro1's figure covers 24 hours, not a cumulative total.

The totals mix company data, creator content and contributor work: Outlier's figures pay individuals for work, and Troveo's first total goes to creators. Pangea.ai is a marketplace and Shfa.ai a medical data program with no published legal entity, and they report what passed through them, not purchases of the data. The company data partnership programs behind these totals are described one by one.

When One Company Publishes Two Totals

Four companies give two different totals on their own pages or filings on the same date, and this page shows both rather than choosing one.

• Mercor, feed checked 2026-11-09: USD 50,000,000 "paid to companies for their data" in the figures block against USD 17,000,000 in the stats block, and USD 2,000,000 paid to referrers "in 2026" against USD 90,000. The meaning of the stats block is not documented.
• Troveo, re-read 2026-11-10: USD 50M+ paid to creators on the content-owners page against USD 20M+ paid to licensors on the business-data page, neither dated.
• Outlier, checked 2026-11-09: USD 500M+ on the home page against USD 600M+ on the about page.
• Wiley, earnings materials of 2024-09-05: USD 16 million as the first-quarter contribution of its second project in the release against USD 17 million in the presentation.

Values that change over time, as opposed to values that disagree on one day, belong in the buyer program changelog.

What a Paid-Out Total Leaves Undefined

A paid-out total answers how much, but not to whom, over what period or in what form, and most published totals leave all three open. Six questions recur, each raised by a case on this page:

• Who are the recipients? Handshake's "partners" is not defined.
• Over what period? Troveo dates neither total, and Pangea's USD 6 million (checked 2026-11-09) covers "the last 60 days" on an undated page.
• Gross or net? Mercor counts referrer payouts in a separate figure.
• Cash or equity? Shutterstock reports equity instruments received as consideration, and Pathos may pay up to 50% of its Tempus fees in preferred stock.
• Company data or contributor work? Outlier's totals pay individuals for tasks.
• One deal or many? Pangea's largest deal, USD 300,000 for 326 repositories from one company (checked 2026-11-09), sits next to a total over 11,069 repositories.

A figure that leaves these open cannot be compared with another, which is why the table has no total row.

Advertised Ranges per Company Against the Record

Advertised ranges per company start at USD 10,000 and reach USD 4 million or more (checked 2026-11-09), and the labels on the pages themselves say these are estimates, not offers or payments. Mercor's listings show USD 10,000 to 1,500,000 across 25 listings, and micro1 states payout tiers of USD 100K to 500K, 500K to 3M and 3M+ per company.

Handshake's headline reads "Earn $100K to $4M+" while its estimator on the same page says the data "could be worth up to $100,000", and its license page says payouts vary with "the outcome of our due diligence process and your acceptance of our data licensing terms and conditions" (re-read 2026-11-10). Each program's advertised range sits in the terms table, dated, under AI data partnership payouts and program terms compared. Three qualifier words recur on the pages:

• Scale AI: "illustrative value", for USD 10K to 1M+ per data partnership (checked 2026-11-09).
• Mode: "Indicative only", for USD 100K to 5M (checked 2026-11-09).
• Troveo: "ILLUSTRATIVE EXAMPLE" for the payout card and "directional" for the estimator.

Against the record, the one intermediary that describes completed deals, SimpleClosure, states tens of thousands to hundreds of thousands of US dollars for startup assets (2026-09-03). The one named seller reports about USD 25,000 net for a codebase with about USD 15,000 pending for workspace data, which is his own estimate (MIT Technology Review, 2026-10-13).

Sellers of data from companies that have closed sit at that end of the record, and the options are set out under selling data from a shut-down or defunct company.

Amounts Confirmed by Sellers, Filings and Courts

Confirmed amounts come from three places: a seller's own statement, a seller's securities filing, and a court record, and only the first two show money actually received. The table groups them by source type, as of 2026-11-09.

A filing confirms receipt but rarely names the buyer, as Reddit's and Wiley's filings show. A contract value is not cash either, since Tempus's Neuro order form of 2026-10-22 is "up to" USD 55.0 million. Contract values sit in the tracker of AI data licensing deals with parties, dates, values and sources.

Court amounts form their own class. The USD 10,000,000 successful bid and USD 7,500,000 alternate bid at the Spirit auction of 2026-08-14 (not approved as of 2026-11-09) are not payments until a sale order is entered. The Bartz initial distribution pays rightsholders under a settlement, not sellers under a license, and USD 2,203.56 is the initial figure per Claimed Work, not the final total.

Payouts to Creators and Contributors Run Through Separate Ledgers

Payouts to photographers, voice owners and contributors are reported by the platforms that pay them, and only one of those reports sits in a securities filing. Getty Images' 10-K, filed 2026-03-16, states royalties of over USD 220 million to content contributors in 2025 across all products, and says it compensates contributors for use of their content as training data; the filing gives no rate. Getty's chapter 11 petition of 2026-10-21 (D. Del.) came with first-day motions to pay contributors in the ordinary course.

• Shutterstock announced its Contributor Fund on 2022-10-25 and says earnings from the OpenAI datasets first reached contributors in Q4 2022; no total is published.
• ElevenLabs' blog headline of 2024-10-28 reads "We crossed $1M in Voice Library Payouts", a company claim with no period or recipient count.
• Outlier's totals, in the table above, pay contributors for work and are not data sales.

Hourly rates for contributor work are covered under pay for AI training work.

Using Reported Payouts in a Decision: Benefits, Risks, Downsides and Alternatives

Reported payouts show what payers want sellers to expect, and they are useful only once each figure is read at its evidence level.

Benefits of reading reported payouts:

• Dated advertised ranges give a seller an anchor to ask about.
• Filings show real deal scale for content holders.

Risks of relying on them:

• Totals with undefined scope, such as Handshake's "partners", cannot be compared.
• Contradictions on one company's own pages, as at Mercor and Troveo, mean no single figure is safe to quote.
• A court bid read as a sale, as with Spirit Airlines, overstates what a seller has received.

Downsides of the evidence itself:

• Confirmed amounts are few and mostly come from large public companies.
• The one named small-seller figure, about USD 25,000 (2026-10-13), is far below the advertised minimums, and this does not imply that any buyer underpaid.

Alternatives to relying on a published figure:

• Ask the buyer for references or a written offer before relying on a range.
• Consider a non-exclusive license or a revenue share, covered under when and how AI data buyers pay.
• Request an independent introduction to compare written terms.
• Not selling or licensing the data at all is a valid outcome.

A payout report is one input; what company data is worth to AI companies depends on rights, exclusivity, volume and the buyer's use as well.

Where Payout Reports Connect to Prices, Terms, Deals and Payment

Payout reports are one layer of the market section; prices, program terms, deal values, filing revenue and payment mechanics each have their own page.

• The AI training data market: size, prices, payouts and history.
• AI training data prices per unit: per-unit prices, dated.
• When and how AI data buyers pay: payment triggers, timing and clawbacks.
• Red flags in data licensing offers: checks for offers that quote large payouts.
• AI data licensing revenue in company filings: the full filing series behind the receipt rows above.
• Bartz v. Anthropic settlement: claims, rulings and distribution status.
• Creators selling content for AI training: the creator side.
• ElevenLabs Voice Library payouts: the voice-owner program.
• Spirit Airlines' data archive sale: the auction and its docket.

How These Payout Figures Are Re-checked and Corrected

Buyer pages are re-read monthly and before any change to the tables, filings each quarter and court dockets before each update. Every changed value keeps its old check date in the corrections log, and the next scheduled re-check is the Spirit Airlines hearing of 2026-11-17.

Frequently Asked Questions

How much did companies get paid for data?

On the record, one named founder reports about USD 25,000 for a codebase, while filings show eight-figure amounts received and nine-figure contract values at large publishers and data holders; program totals are the payers' own claims. Tempus AI states USD 50 million upfront paid as of April 2025, and Reddit states a USD 203.0 million aggregate contract value in its S-1 of 2024-02-22.

Has any company confirmed a payout from micro1, Mercor or Handshake?

No seller has publicly confirmed an amount received under any of the three programs; their totals and ranges are their own claims, checked November 9, 2026. micro1 states USD 5.8 million in 24 hours, Mercor USD 50,000,000 and USD 17,000,000 in two blocks, and Handshake USD 250 million.

Are paid-out totals on buyer websites audited?

Nothing on the pages says so; only Getty's contributor royalty figure appears in an annual report filed with the SEC. The other totals are statements on company pages or posts.

Did Google pay USD 10 million for Spirit Airlines' data in 2026?

No: Google LLC's USD 10,000,000 is the successful bid at the 2026-08-14 auction (Dkt. 1463), and the court had not approved the sale as of November 9, 2026; the hearing is set for November 17, 2026. A bid is not a payment, and the rules for approval are explained under US bankruptcy data sales.