Major news companies are making more commercial deals with AI firms—but “selling out” is a judgment, not a precise description. The arrangements range from licensing archives and live news to using AI tools inside a newsroom. Some publishers are also suing over other uses of their work. Whether these deals help journalism depends on what rights they grant, what protections they include, and whether the money and referrals offset any loss of audience.
What “selling out to AI” can mean
The phrase can describe several different choices, but they are not interchangeable. A publisher may license its archive for model training, let a chatbot retrieve and summarize current reporting, buy enterprise software for staff, or build a reader-facing AI product. It may also allow AI-assisted work into its publishing process. Each involves different rights, risks and commercial terms.
- Content licensing: granting specified rights to use journalism, which may cover archives, current articles, training, retrieval or display.
- Distribution: allowing an AI product to quote, summarize or link to reporting when answering users.
- Newsroom adoption: using AI for tasks such as translation, research assistance, data analysis or archive search. This does not by itself grant the vendor rights to publish or train on the publisher’s journalism.
- Automation: using AI to produce or modify material for publication, a separate editorial decision with its own oversight questions.
A public “partnership” announcement does not necessarily disclose which of these applies. The contract may also contain terms—such as payment, audit rights and termination conditions—that are not public.
The deal wave is broad, but its terms vary
The commercial trend is not limited to a few early agreements. On January 15, 2025, OpenAI said it had partnerships or programs involving nearly 20 media organizations and more than 160 outlets or content brands. Meta’s announcement, updated July 27, 2026, listed a broad group of partners for real-time news in Meta AI. Those counts show reach, not that every partner granted the same rights or received the same terms.
#1 Best Overall
| Publisher or group | Partner and public description | What the announcement establishes—and what it does not |
|---|---|---|
| Associated Press | OpenAI; licensing access to part of AP’s archive. AP also announced a Google arrangement to deliver up-to-date news through Gemini. | AP’s OpenAI deal terms were not disclosed. Public descriptions do not establish identical display, attribution or linking arrangements for the Google product. AP coverage of the Google agreement. |
| Axel Springer | OpenAI; selected summaries from brands including POLITICO, Business Insider, BILD and WELT, along with training use and product development. | The announcement describes both display and training-related uses; it is not evidence that every article is freely reproduced. OpenAI’s announcement. |
| Financial Times | OpenAI; attributed summaries, quotes and links, model improvement, and joint product work. | The FT also became an OpenAI enterprise customer. The public announcement does not supply all contract economics. OpenAI’s announcement. |
| News Corp | OpenAI; current and archived content from named publications, including The Wall Street Journal, Barron’s, MarketWatch, The Times, The Sun and The Australian. | The agreement covers specified News Corp publications, not necessarily every business in the group. OpenAI’s announcement. |
| Condé Nast | OpenAI; content from brands including Vogue, The New Yorker, Wired, Vanity Fair, GQ and Bon Appétit in OpenAI products. | The announcement describes content integration and discovery; it should not automatically be read as unrestricted training access. OpenAI’s announcement. |
| Guardian Media Group | OpenAI; Guardian journalism in ChatGPT, with attributed summaries and links, plus ChatGPT Enterprise deployment. | Content use and the Guardian’s internal software adoption are distinct parts of the relationship. OpenAI’s announcement. |
| Vox Media | OpenAI; a publisher partnership involving news content. | The cited announcement does not publicly specify the complete rights or commercial terms. OpenAI’s overview of its news partnerships. |
| Reuters | OpenAI and Meta-related products; licensing and distribution partnerships are described in the available material. | Reuters is a news agency; it should not be conflated with Thomson Reuters’ legal, tax and professional-information businesses. Confirm the particular product and rights before calling an arrangement a training deal. |
| TIME | OpenAI and other AI-related programs; publisher partnership and content integration. | The available descriptions do not establish that all arrangements grant model-training rights. The particular product and rights matter. |
| CNN, Fox News, USA TODAY, People Inc. and other groups | Meta; real-time content in Meta AI. | Meta identifies these as content partners, but public information does not establish identical payment, training rights or contract terms across the group. Meta’s announcement. |
The table captures publicly described purposes, not complete contracts. A license for retrieval or display is not automatically permission to train a model, and a newsroom software deployment is not the same as either.
News Corp shows why “sellout” is too simple
News Corp has made agreements with both OpenAI and Meta, while its chief executive has said the company will negotiate where possible and sue when its content is used without permission. Its OpenAI partnership covers current and archived material from named titles. Separately, The Guardian reported on March 4, 2026, that the Meta agreement covering U.S. and U.K. content could be worth up to $50 million annually over three years. That is a reported ceiling, not a confirmed guaranteed annual payment; Meta’s own announcement names News Corp as a partner but does not state the amount. The Guardian’s report.
The figure may sound large, but without the contract, relevant revenue comparisons, the covered titles and uses, and any traffic effects, it cannot establish whether the arrangement is economically favorable. Nor does a group-level agreement mean that every newsroom, archive or brand has the same contract or policy.
What publishers are licensing—and what they hope to get
Archives and model improvement
Older reporting can be valuable to an AI company seeking reliable material. Some public agreements explicitly mention training or model improvement; others describe content use without spelling out training rights. News Corp’s OpenAI announcement refers to current and archived content, while Axel Springer’s describes both selected content in answers and use of quality content to advance model training. The Financial Times announcement includes attributed content, model improvement and joint product development. Those distinctions matter more than the generic label “AI deal.”
Rank #3
Current reporting in answers
A live or recent-news license can let an AI product retrieve reporting and present summaries, excerpts or links to users. Publishers hope the exposure sends readers to original work. OpenAI says its partnerships are intended to help users discover reporting through summaries, citations and links, but the existence of a citation does not show that users click it or subscribe. OpenAI’s description of its news partnerships.
Tools and product experiments
Some publishers are testing archive search, translation, personalization and AI-assisted discovery, while also adopting enterprise tools for internal work. These may help staff find material or process data, but they do not resolve the separate question of what content an outside company may use. OpenAI’s July 22, 2026 account of newsroom uses—including reporting, verification, translation and business workflows—is the vendor’s description of its own technology, not independent proof of effectiveness. OpenAI’s account.
Revenue and bargaining leverage
Licensing offers publishers a possible new income stream from journalism that AI systems need. The Associated Press has described diversification as part of its business strategy, while its OpenAI financial terms were not disclosed. A deal can also create a negotiating precedent with other AI firms. Whether the revenue is material depends on the amount, duration, covered content, traffic impact and protections—details often absent from public announcements.
Why sign a deal and still sue?
Publishers are not necessarily choosing between permanent opposition and unconditional cooperation. The New York Times has sued OpenAI and Microsoft over alleged unauthorized use of its journalism and, according to Axios, separately reached a multiyear AI licensing arrangement with Amazon in 2025. A publisher can object to one company’s alleged use while licensing a different use to another company. The legal claims remain disputed; a separate license does not decide whether unlicensed training is lawful. Axios’s report.
Different deals may involve different uses, products, safeguards and bargaining positions. A publisher might accept payment for retrieval in a product that links back while contesting scraping or training elsewhere. Litigation and licensing can therefore be parallel strategies rather than opposites.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.The risks publishers and readers should watch
- Traffic substitution: A chatbot may answer a question well enough that a user never visits the original article. A link in the answer does not, by itself, compensate for lost referrals, advertising impressions or subscriptions.
- Unequal bargaining power: Major AI companies have substantial capital and distribution. Publishers facing declining referral traffic may negotiate under pressure, and a one-time deal may not preserve leverage at renewal.
- Opaque economics and rights: Public announcements often omit fees, duration, exclusivity, audit rights, usage reports and termination provisions. A reported maximum should not be treated as a guaranteed payment.
- Errors and context: AI summaries can misstate, omit or combine reporting. Source attribution does not guarantee that the answer is accurate, current, or clear about the difference between reporting and opinion.
- Editorial independence: Payment from a technology company can create questions about perceived conflicts, even without evidence of editorial restrictions. Do not assume censorship; look for documented safeguards and actual editorial policies.
- Labor and workflow changes: Translation or data tools may assist journalists, but their availability alone does not prove job losses. Assess documented staffing or production changes rather than assuming every efficiency tool displaces workers.
- Concentration and unequal access: Large groups have more capacity to negotiate licenses than local or nonprofit outlets. OpenAI’s newsroom programs and grants do not automatically amount to lucrative content licensing.
- Dependence on a vendor: If a publisher builds search, archive access or audience discovery around one provider, switching can become costly and direct relationships with readers may weaken.
The underlying industry problem is sharpened by the growth of search and AI answers that can satisfy users without a click. The Reuters Institute has reported publisher concern about referral declines and zero-click search following Google’s AI Overviews; that concern is not, by itself, a quantified measure of losses caused by any particular licensing agreement. Reuters Institute analysis.
Publishers are also trying to set common rules
On February 26, 2026, major British media organizations including the BBC, Financial Times, Guardian and Sky News backed an initiative known as SPUR to promote common technical standards and licensing frameworks. Its stated aim is rights-cleared access to journalism with publisher control and payment. The initiative is evidence of collective concern and an effort to coordinate; it does not establish that a functioning, enforceable global licensing system already exists. The Guardian press office announcement.
How to judge whether a particular deal is fair
Readers, journalists and investors can use a contract-focused checklist rather than treating a publisher’s participation as proof of surrender or protection:
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →- Rights: Does the agreement cover archive training, current-content retrieval, display, search, fine-tuning, internal data or commercial products beyond chatbots?
- Scope: Which titles, markets, formats and subscriber-only materials are included? Can a publisher exclude sensitive archives or investigations?
- Payment: Is compensation fixed, usage-based or shared revenue? Is there a minimum guarantee, and are figures guaranteed or merely reported as a ceiling?
- Accountability: Are there usage reports, audit rights, correction channels, provenance information and a way to address inaccurate or outdated answers?
- Control: Can the publisher opt out, terminate access or prevent full-article reproduction? What happens at renewal?
- Audience value: Are links prominent, and does the publisher measure click-throughs, repeat visits and subscriptions rather than just counting citations?
- Editorial safeguards: Are human review, source protection, correction practices and disclosure of AI assistance clearly governed?
- Exit and dependence: Can the publisher leave without rebuilding critical systems or losing access to its own archive and audience data?
Until contract terms and results are clearer, the most accurate description is commercial accommodation under pressure, not a blanket surrender. These deals may create a second revenue stream and new routes to readers—or help platforms answer questions without sending readers to journalism. Which outcome prevails depends on rights, economics, safeguards and measurable audience behavior.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
