Market
Data is licensed for money. Here is the public record.
Public data-licensing deals
None of these companies is a Boxtrunk buyer or partner.
These are other companies' deals. They show that data is licensed for money, not what any business's records are worth.
Content and publisher deals
News, books, images, code answers and forum posts, licensed by the publishers and platforms that hold them.
Reddit → Google
Access to Reddit's Data API (“real-time, structured, unique content”), plus a Vertex AI cloud partnership · 2024
Not disclosed · Google blog, Feb 22, 2024
“worth about $60 million per year” (reported by Reuters, according to one of three anonymous sources) · Reuters via Yahoo, Feb 22, 2024
Reddit → Unnamed licensees
Data licensing arrangements · 2024
“In January 2024, we entered into certain data licensing arrangements with an aggregate contract value of $203.0 million and terms ranging from two to three years” (stated in Reddit's S-1 filing) · TechCrunch, Feb 22, 2024
“We expect a minimum of $66.4 million of revenue to be recognized during the year ending December 31, 2024” (stated in Reddit's S-1 filing) · TechCrunch, Feb 22, 2024
The filing does not name the licensees.
Reddit → OpenAI
Access to Reddit's Data API; OpenAI also becomes a Reddit ad partner · 2024
Not disclosed · Reddit (Upvoted blog), May 16, 2024
News Corp → OpenAI
Current and archived content from WSJ, NY Post, and UK and Australian titles · 2024
Not disclosed · Variety via Yahoo, May 22, 2024
“the deal could be worth more than $250 million over five years.” (reported by The Wall Street Journal, as cited by Variety) · Variety via Yahoo, May 22, 2024
Shutterstock → OpenAI
“Shutterstock's image, video and music libraries and associated metadata” · 2023
“new six-year agreement” (stated by Shutterstock) · Shutterstock press release via AAP, Jul 11, 2023
No financial terms disclosed · Shutterstock press release via AAP, Jul 11, 2023
Shutterstock → Big Tech licensees (aggregate)
AI licensing of its content library · 2023
“earned $104 million in annual revenue from AI licensing agreements in 2023.” (stated by Shutterstock CEO Paul Hennessy) · Business Insider via AOL, updated Dec 16, 2024
“initially ranged from $25 million to $50 million each”, and “most were later expanded” (stated by Shutterstock's CFO, to Reuters) · Reuters via China Daily HK, Apr 6, 2024
Stack Overflow → Google Cloud
Stack Overflow knowledge via the new OverflowAPI, for Gemini for Google Cloud · 2024
Not disclosed · Stack Overflow press release, Feb 29, 2024
Stack Overflow → OpenAI
OverflowAPI access, to improve OpenAI models and surface content in ChatGPT · 2024
Not disclosed · Stack Overflow press release, May 6, 2024
Axel Springer → OpenAI
Politico, Business Insider, Bild and Welt content for training and for ChatGPT summaries · 2023
“tens of millions of euros” per year (reported by Financial Times, as cited by The Decoder) · The Decoder, Dec 19, 2023
“tens of millions of euros” over a three-year term (reported by Bloomberg, as cited by Euronews) · Euronews, Dec 14, 2023
The companies did not disclose terms, and the two reports disagree on whether the figure is per year or over three years.
Associated Press → OpenAI
“part of AP's text archive,” with “AP news stories going back to 1985” · 2023
“Financial terms of the deal were not disclosed.” · AP via PBS NewsHour, Jul 13, 2023
Associated Press → Google
Real-time news feed for the Gemini app · 2025
“Neither company has disclosed how much Google will pay AP” · AP via PBS NewsHour, Jan 15, 2025
Financial Times → OpenAI
A “strategic partnership and licensing agreement”; FT journalism used in ChatGPT and to “help improve its models' usefulness” · 2024
Not disclosed · Financial Times press release, Apr 29, 2024
Vox Media and The Atlantic → OpenAI
Content used “to train its AI models and be shared inside of ChatGPT” · 2024
“Terms for the deals with The Atlantic and Vox Media weren't disclosed.” · The Verge, May 29, 2024
Condé Nast → OpenAI
Content from The New Yorker, Vogue, GQ, Vanity Fair, Bon Appétit and others for ChatGPT and SearchGPT · 2024
“did not disclose the terms of the deal” · Al Jazeera, Aug 21, 2024
Dotdash Meredith (IAC) → OpenAI
Content licensing · 2024
“at least $16 million per year” (reported by Adweek, from its review of public financial documents, as cited by Engadget) · Engadget, Nov 19, 2024
The New York Times → Amazon
News, NYT Cooking and The Athletic, for AI training and Alexa summaries · 2025
Not disclosed · TheWrap, Jul 30, 2025
“between $20 million to $25 million per year” (reported by The Wall Street Journal, as cited by TheWrap) · TheWrap, Jul 30, 2025
Reuters → Meta
News content so Meta AI can answer news queries · 2024
Multi-year; financial terms not public · Engadget, Oct 25, 2024
Whether the deal covers training is “not disclosed”. · Engadget, Oct 25, 2024
Informa / Taylor & Francis → Microsoft
Access to “advanced learning content and data” · 2024
“an initial data access fee of around $10 million along with recurring payments over the next three years” (reported by interactive investor) · interactive investor, May 8, 2024
Wiley → Two undisclosed large tech companies
Previously published learning and research content for model training · 2024
“a $23 million AI deal with an undisclosed large tech company” (FY2024) (reported by Publishers Weekly) · Publishers Weekly, Sep 9, 2024 print issue
“another agreement with a second tech company worth $21 million” (implemented Q1 FY2025) (reported by Publishers Weekly) · Publishers Weekly, Sep 9, 2024 print issue
Two separate deals in different fiscal years.
HarperCollins → Microsoft
Three-year AI training license; authors free to accept or decline · 2024
“$5,000 per book”, with authors receiving “half of that amount” (reported by The Decoder) · The Decoder, Feb 8, 2025
Internal business and workplace data
Internal business data from a company in bankruptcy and from companies that have shut down. These are the newest and least settled examples.
Spirit Airlines (in bankruptcy) → Google (sale)
Internal business data: “employee emails, Microsoft Teams messages, spreadsheets, and calendars,” plus marketing, productivity and operations data · 2026
Not court-approved as of Oct 10, 2026. The court hearing is set for Nov. 6, 2026. · Miami Herald via Yahoo, Oct 10, 2026
“for $10 million” (reported by Reuters) · Reuters via The Star, Aug 18, 2026
“The data will be de-identified before the sale is complete” (reported by Reuters) · Reuters via The Star, Aug 18, 2026
A court-appointed privacy ombudsman recommended on Oct. 5, 2026 that the sale go forward. · Miami Herald via Yahoo, Oct 10, 2026
“Data on passengers aren't part of the proposed sale.” · Miami Herald via Yahoo, Oct 10, 2026
Defunct startups (via SimpleClosure) → AI developers (not named) (sale)
Source code and workplace data (Slack messages, emails, documents, workflows) for AI training “gyms” · 2026
“SimpleClosure has processed nearly 100 of these deals” (reported by Forbes, as cited by Gizmodo) · Gizmodo, Apr 17, 2026
Investment in a training-data company
An investment in a company that labels and supplies training data. It is not a purchase of records.
Meta → Scale AI (investment)
Investment in a data-labeling and training-data vendor · 2025
“$14.3 billion” for “a 49% stake” (reported by CNBC) · CNBC via NBC Chicago, Jun 12, 2025
Meta has no voting power. · CNBC via NBC Chicago, Jun 12, 2025
Estimates of the AI training-dataset market
Each firm defines this market differently, and some include annotation services and software, so these figures are not comparable with each other. They are shown side by side and never averaged.
Grand View Research
2025 → 2033
“expected to reach USD 16,320 million by 2033, registering a CAGR of 22.6% from 2026 to 2033” (analyst estimate) · Grand View Research press release, March 2026
Base year and forecast period: 2025 → 2033. Market scope as defined by Grand View Research.
MarketsandMarkets
2024 → 2029
“slated to expand from USD 2.82 billion in 2024 to USD 9.58 billion by the year 2029 at a robust CAGR of 27.7%” (analyst estimate) · MarketsandMarkets press release, Oct 24, 2024
Base year and forecast period: 2024 → 2029. Market scope as defined by MarketsandMarkets.
Verified Market Research
2023 → 2031
“valued at USD 1555.58 Million in 2023 and is expected to reach USD 7564.52 Million by the end of the forecast period”, at a “CAGR of 21.86% from 2024 to 2031” (analyst estimate) · Verified Market Research via PR Newswire, Apr 14, 2025
Base year and forecast period: 2023 → 2031. Market scope as defined by Verified Market Research.
What these deals say about operational records
Most of the deals above license news, books, images, and forum posts: media and platform content, not work orders or service logs.
What they show is what buyers pay for: data that is hard to find elsewhere, cleared for use, and well documented. Those are the things the assessment checks in your records.
The closest examples to operating records are the newest and least settled: the Spirit Airlines sale is not court-approved, and the defunct-startup sales rest on one report. None of these deals is a price for operational records.
What makes operational records licensable
Records that can qualify
Work orders
What was asked for and what was done
Service logs
Every visit, in order, over years
Technician notes
What the person on site actually found
Support histories
Questions, answers, and follow-ups
Procedures
How your team is trained to do the work
Quality records
Callbacks, rework, and inspections
Estimates
What was quoted, and what was approved
What makes records strong
- Rights
- The records are clearly yours to license under your contracts and your customer terms.
- Consent
- Your notices and agreements with customers and staff allow the records to be used this way.
- Completeness
- Each record shows the request, the work done, and the result, with few gaps.
- Outcomes recorded
- The records show how each job turned out, including callbacks and rework.
- History length
- The records go back years, without long breaks.
- De-identification
- Names, addresses, and other identifiers can be removed or replaced in the licensing copy, and the result is checked before approval.
Health records, consumer financial records, records about children, and biometric data come with specific legal restrictions.
If your records include any of them, we review your situation case by case, on its own, and the answer may be no.
What shapes a license's terms
- Exclusivity
- Whether one buyer alone may use the records, and for how long.
- Specialization
- How specialized the work in the records is, and how hard similar records are to find.
- Quality and preparation
- How consistent and well organized the records are, and how much work they need before delivery.
- Freshness and coverage
- How recent the records are, and how much of the work, time, and geography they cover.
- Rights and privacy
- How clear the rights are, and how much personal information has to be removed first.
- License scope
- What the buyer may do with the copy: which uses, for how long, and with which limits.