AI data licensing deals, 2023 to 2026: the sourced list.
Who licensed content or records to which AI company, when, for how much, and where that is on the record. Last checked 27 October 2026. If your company owns workplace records, Allied arranges a direct license with an AI lab.
How this list is built
A deal is listed only if one side announced it, a securities filing or court sale discloses it, or a major outlet reported it. Every row links to the page we read when we checked it. The date at the top is the last time every source was reopened.
“Not disclosed” means a source says the terms were not released. “No figure published” means the announcement gives no price and we found no credible report of one. “Reported” means a news outlet gave a figure the companies did not confirm; each such row names the outlet. Figures from Reddit, IAC, Informa and Wiley come from their own filings or results.
Some deals license archives for model training. Others pay for content shown inside a chatbot or search answer. The “What was licensed” column says which when the source does.
To cite this page: Allied Advantage Assets, “AI data licensing deals, 2023 to 2026,” alliedassets.ai/ai-data-licensing-deals/, last checked 27 October 2026.
The deals, oldest first
Twenty-five deals. Numbers in each row point to the sources at the foot of the page.
| Licensor | AI company | Announced | Value | What was licensed | Source |
|---|---|---|---|---|---|
| Shutterstock | OpenAI | 11 July 2023 | Not disclosed1 | Six-year license to its image, video and music libraries and their metadata, as training data | Shutterstock release1 |
| Associated Press | OpenAI | 13 July 2023 | Not disclosed2 | Part of AP’s text archive, with stories back to 1985 | AP2 |
| Axel Springer | OpenAI | 13 December 2023 | Not disclosed3; later reported as tens of millions of euros4 | Articles from Politico, Business Insider and other titles, for training and for summaries in ChatGPT | TechCrunch3 |
| Several licensees, not named in the filing | January 2024 | $203.0 million in total, over terms of two to three years5 | API access to Reddit data plus quarterly transfers | Reddit Form S-15 | |
| 22 February 2024 | About $60 million, per AP6 | Reddit posts through its Data API, for training Google’s models | AP6 | ||
| Financial Times | OpenAI | 29 April 2024 | No figure published7 | FT journalism, to improve OpenAI’s models and for attributed answers in ChatGPT | FT release7 |
| Stack Overflow | OpenAI | 6 May 2024 | No figure published8 | Developer questions and answers through OverflowAPI | Stack Overflow release8 |
| Dotdash Meredith | OpenAI | May 20249 | At least about $16 million a year, fixed part only, per Adweek’s reading of IAC filings10 | Content from its publishing brands | IAC 10-Q9, Adweek10 |
| Taylor & Francis (Informa) | Microsoft | 8 May 2024 | $10m+ initial data access fee, then recurring payments for 2025, 2026 and 202711 | Non-exclusive access to Advanced Learning content and data | Informa market update11 |
| OpenAI | 16 May 2024 | No figure published12 | Real-time Reddit content through its Data API, for ChatGPT and new products | Reddit announcement12 | |
| News Corp | OpenAI | 22 May 2024 | More than $250 million over five years, reported4 | Current and archived content from The Wall Street Journal, Barron’s, MarketWatch, the New York Post and other titles | News Corp release13, TechCrunch4 |
| Vox Media | OpenAI | 29 May 2024 | Not disclosed4 | Current content and its full archive, to train ChatGPT and other models | TechCrunch4 |
| The Atlantic | OpenAI | May 2024 | Not disclosed4 | A deal similar to Vox Media’s. The Atlantic’s union said neither side would answer questions about its terms | TechCrunch4 |
| Taylor & Francis (Informa) | Second AI partner, not named | July 2024 | Took Informa’s AI partnership revenue to $75m+ for 202414 | Partnership and data access agreement | Informa half-year results14 |
| Condé Nast | OpenAI | August 2024 | Not disclosed15 | Content from The New Yorker, Vogue, GQ, Vanity Fair, Bon Appétit and other brands | Al Jazeera15 |
| Wiley | Large tech companies, not named | Fiscal years to April 2024, 2025 and 2026 | AI licensing revenue of $23 million in fiscal 2024 and $40 million in fiscal 202516, then $49.1 million in fiscal 202617 | Content licensing; a third large tech company signed in early 2025 | Wiley 8-K16, Wiley 10-K17 |
| Associated Press | 15 January 2025 | Not disclosed18 | A real-time news feed for the Gemini app | AP18 | |
| AFP | Mistral AI | 16 January 2025 | No figure published19 | AFP’s newswire, about 2,300 stories a day in six languages, for the Le Chat assistant | Mistral announcement19 |
| The New York Times | Amazon | May 2025 | $20 million to $25 million a year, as first reported by The Wall Street Journal20 | News, NYT Cooking and The Athletic, for training Amazon’s models and for summaries in Alexa | TheWrap20 |
| Getty Images | Perplexity | 31 October 2025 | No figure published21 | Display of Getty images in Perplexity’s search tools; a display license, not training | Getty Images release21 |
| CNN, Fox News, People Inc, USA Today, Le Monde and others | Meta | 5 December 2025 | Not disclosed22 | Real-time news for answers in Meta AI | Reuters22 |
| Wikimedia Foundation | Amazon, Meta, Microsoft, Mistral AI, Perplexity | 15 January 2026 | No figure published23 | Wikipedia and other Wikimedia project data through Wikimedia Enterprise | Wikimedia Enterprise23 |
| News Corp | Meta | March 2026 | Up to $50 million a year over three years, reported by The Wall Street Journal24 | The Wall Street Journal and other US and UK brands, for Meta AI answers and model training | Engadget24 |
| Nine Entertainment | Microsoft | 3 July 2026 | No figure published25 | Nine’s mastheads in answers from Microsoft Copilot | Microsoft release25 |
| Spirit Airlines (in bankruptcy) | August 2026 | $10 million26 | Employee emails, Microsoft Teams messages, spreadsheets, calendars and operations data. Sold at auction, not licensed | Reuters26, MIT Technology Review27 |
What the list shows
Almost every row is published content. Twenty-four of the 25 license news, books, images, reference pages or public forum posts. One, the Spirit Airlines sale, covers a company’s internal work records.
Most prices stay private. Fourteen rows carry no figure. Of the eleven that do, five come from the companies’ own filings or results, five are press reports the companies never confirmed, and one is a court-supervised auction.
OpenAI is the buyer in eleven rows, all signed in 2023 and 2024. From 2025 the buyers spread out to Google, Amazon, Meta, Microsoft, Mistral AI and Perplexity.
The purpose is moving from archives to live answers. AP now feeds real-time news to Google’s Gemini app,18 Meta’s publisher deals supply news for Meta AI,22 and Nine’s journalism appears in Copilot answers.25
Licensing has become a repeat line of revenue for some publishers. Wiley’s AI license revenue reached $49.1 million in its fiscal 2026, and $19.4 million of that came from content Wiley had itself licensed from other publishers.17
What this means if you own workplace records
The list is mostly publishers because publishers announce. A media company benefits when readers and advertisers hear that AI labs pay for its work. A company licensing its Slack or Teams history, Zendesk tickets, Salesforce records or GitHub code has no such reason, so most of that market never reaches the press.
The one large workplace-records sale on the record came through a bankruptcy court. Google outbid the AI data company Mercor for Spirit Airlines’ archive: 100 million emails, 500 million Teams messages and 176,000 employee records, for $10 million.27
Court sales are public, and they draw objections. Spirit’s flight attendants’ union objected to the sale on privacy grounds.27 The Bankruptcy Code bars a trustee from selling personally identifiable information against the debtor’s own privacy policy unless the sale fits the policy or a consumer privacy ombudsman is appointed first.28 The statute defines that information by reference to products or services bought for personal, family, or household purposes.29 On its face, that protects customers more than staff.
Smaller sales happen when startups close. Wind-down firms now sell code and workspace data to AI labs. One of them, Sunset, told MIT Technology Review that it values a client’s data by head count, years in operation, volume and how the team used its tools. The same article reports that buyers quote very different numbers and that sellers often cannot see who is buying.27
No operating company has announced a license of its own email, chat or CRM history. That does not mean none exist. It means none are public yet, and this list will add the first one that is. Which kinds of buyer are active is covered in who buys AI training data; what sets the fee is in what company data is worth.
Licensing your records without a court sale
A company that is still running does not need an auction. It can license an anonymized copy of its records to an AI lab and keep the originals. Allied arranges it, and never sees the data.
What happens: a short questionnaire, then a discovery call about your systems and policies. A written offer naming the lab follows in one to two weeks. Names, contact details and customer identities are removed before anything leaves, and you approve a sample. Payment arrives by wire 30 to 60 days after the lab accepts delivery.
The license is non-exclusive. The usual fit is 30 or more office employees with more than a year of history. A typical license runs $100K–$2M. The clauses are explained in AI data licensing agreements and the legal checks in is it legal to sell company data. Request an offer, see the six systems labs license, or read how Allied works.
Questions about the list
Which AI data licensing deal is the largest on record?
Among deals with a public figure, News Corp’s five-year deal with OpenAI, going by press reports. Neither company confirmed the number. The table gives the figure and its source.
Why do most deals have no price?
Prices surface only when a public company has to report them, a court supervises a sale, or a reporter learns them. Private licensors and AI labs have no duty to publish terms.
Is a data license the same as selling data?
No. A license lets the AI company use a copy for agreed purposes. The owner keeps the original and can license it again unless the deal is exclusive. The Spirit Airlines records were sold outright at a bankruptcy auction.
How often is this list updated?
The date at the top of the page is the last check. Each check reopens every source and adds deals made public since the one before.
Are deals arranged through Allied on this list?
No. The list covers only deals that are on the public record.
Sources
- Shutterstock, “Shutterstock Expands Partnership with OpenAI, Signs New Six-Year Agreement to Provide High-Quality Training Data,” July 11, 2023 (press release via AAP)
- Associated Press via PBS NewsHour, “ChatGPT creator OpenAI signs deal with AP to license news stories,” July 13, 2023
- TechCrunch, “OpenAI inks deal with Axel Springer on licensing news for model training,” December 13, 2023
- TechCrunch, “‘What’s in it for us?’ journalists ask as publications sign content deals with AI firms,” June 22, 2024
- Reddit, Inc., Form S-1 registration statement, February 22, 2024
- Associated Press via CBS News, “Google strikes $60 million deal with Reddit, allowing search giant to train AI models on human posts,” February 23, 2024
- Financial Times, “Financial Times announces strategic partnership with OpenAI,” April 29, 2024
- Stack Overflow, “Stack Overflow and OpenAI Partner to Strengthen the World’s Most Popular Large Language Models,” May 6, 2024
- IAC Inc., Form 10-Q for the quarter ended September 30, 2024
- Adweek, “OpenAI Is Paying Dotdash Meredith At Least $16 Million to License Its Content,” November 2024
- Informa PLC, market update, May 8, 2024
- Reddit, “Reddit and OpenAI Build Partnership,” May 16, 2024
- News Corp, “News Corp and OpenAI Sign Landmark Multi-Year Global Partnership,” May 22, 2024
- Informa PLC, 2024 half-year results, July 24, 2024
- Al Jazeera, “OpenAI strikes deal to use content from The New Yorker, Vogue, Vanity Fair,” August 21, 2024
- John Wiley & Sons, Form 8-K, Exhibit 99.1, fiscal 2025 results, June 17, 2025
- John Wiley & Sons, Form 10-K for the fiscal year ended April 30, 2026
- Associated Press via PBS NewsHour, “Google signs deal with AP to deliver up-to-date news through its Gemini AI chatbot,” January 15, 2025
- Mistral AI, “Purr-fectly informed” (AFP partnership), January 16, 2025
- TheWrap, “New York Times Seals $20 Million AI Deal With Amazon,” July 30, 2025 (reporting The Wall Street Journal)
- Getty Images, “Getty Images and Perplexity strike multi-year image partnership,” October 31, 2025
- Reuters via BNN Bloomberg, “Meta strikes multiple AI deals with news publishers,” December 5, 2025
- Wikimedia Enterprise, “Announcing New Wikimedia Enterprise Partners for Wikipedia’s 25th Birthday,” January 15, 2026
- Engadget, “Meta signs a multimillion-dollar AI licensing deal with News Corp,” March 3, 2026 (reporting The Wall Street Journal)
- Microsoft Source Asia, “Nine and Microsoft announce Australian-first AI agreement for news media content,” July 3, 2026
- Reuters via iTnews, “Google to buy Spirit Airlines’ business data for US$10 million,” August 19, 2026
- MIT Technology Review, “AI labs are buying data from dead startups to train their models,” October 13, 2026
- 11 U.S.C. § 363(b)(1), via Cornell Legal Information Institute
- 11 U.S.C. § 101(41A), via Cornell Legal Information Institute
General information, not legal advice. Values are as disclosed or reported on the date checked.
Find out what an AI lab would pay for your records.
A few questions. About two minutes. How Allied works.