The release describes Guardian reporting and archive journalism as a news source within ChatGPT, with attributed short summaries and article extracts, and it does not state training. This page keeps the two records apart within the wider picture of AI training data licensing and shows who said what, and when.
Fact-check note: the Guardian's release and Google's two posts were re-read on 2026-11-17; the approach is described in how we source, date and check every figure.
The Guardian's AI agreements at a glance: counterparties, dates and scope
The public record holds two licensing records for the Guardian and one related litigation row: OpenAI (2025-02-14), Google (2025-12-10) and the Cohere suit.
The table lists them by date, with the party that stated each term (as of 2026-11-17).
| Announced | Counterparty | Access and data | Use as stated | Value as stated (stated by) | Source |
|---|---|---|---|---|---|
| 2025-02-14 | OpenAI | Guardian reporting and archive journalism as a news source in ChatGPT; attributed short summaries and article extracts; ChatGPT Enterprise to be rolled out inside the Guardian | display; training not stated | not disclosed | theguardian.com (GNM press office) |
| 2025-12-10 | Google LLC | named participant in Google's News AI pilot | enhanced news content for AI features in Google News; training not stated | not disclosed per publisher; Google states it pays participating publications | blog.google |
| Related, not a license | Cohere Inc. (S.D.N.Y. 1:25-cv-01305-CM-OTW, filed 2025-02-13) | suit by 14 publishers including Guardian News & Media | n/a | n/a | courtlistener.com |
The deal tracker shows the same two licensing rows, both coded as display with the value not disclosed. The Cohere row is a lawsuit, not a license, and no statement ties it to either agreement.
The two licensing records differ in who speaks. The OpenAI record rests on the Guardian's own release, which carries quotes from an executive of each side.
The Google record rests on Google's posts alone: the pilot names The Guardian, and no Guardian statement on it is recorded in this tracker. Both rows therefore keep their stating party in the table, and neither is read as a statement by the other side.
OpenAI and Guardian Media Group: reporting and archive in ChatGPT, with attributed summaries and extracts
Guardian Media Group announced on 2025-02-14 that Guardian reporting and archive journalism will be available as a news source within ChatGPT, alongside attributed short summaries and article extracts.
The release adds that "the Guardian will also roll out ChatGPT Enterprise to develop new products, features and tools". The Chief Financial and Operating Officer of Guardian Media Group, Keith Underwood, says in the release that the partnership "reflects the intellectual property rights and value associated with our award-winning journalism".
The Chief Operating Officer of OpenAI, Brad Lightcap, says it "enriches the ChatGPT experience by providing access to relevant, real-time news content" and is "part of our overall strategy to help publishers and audiences benefit from advanced AI technology".
The release names the grant as availability in ChatGPT and the publication of summaries and extracts, which is display. It is silent on duration, exclusivity, renewal and deletion, and it gives neither a price nor a payment schedule.
Its boilerplate describes Guardian Media Group's core business as Guardian News & Media, the publisher of theguardian.com, which is the content the release covers. The signing entity is not stated, so the page writes "OpenAI", the brand. OpenAI's other recorded agreements sit on the profile of OpenAI as a data buyer.
Google's news pilot: the Guardian as a named participant
Google's post of 2025-12-10 names The Guardian among the publishers in a new commercial partnership program it is piloting with a range of news publishers globally.
The post also names Der Spiegel, El País, Folha de S. Paulo, Infobae, Kompas, The Times of India, The Washington Examiner and The Washington Post, "among others".
Google's later update of 2026-10-15 says it pays participating news publications for the delivery and use of enhanced content, and counts more than 400 publications. The full participant list and program terms sit on the page about the Google News AI pilot program.
No statement by the Guardian on the pilot is recorded in this tracker, and Google gives no Guardian-specific term. The Guardian's place in the pilot rests on Google's naming alone.
The post of 2025-12-10 describes Google paying for extended display rights and content delivery methods such as APIs. That wording concerns display and delivery. It does not say that the participating content is used to train models, and it does not set out what any single publisher receives.
What the parties state about value and training
Neither Guardian Media Group, OpenAI nor Google states an amount for either arrangement, so this page shows no price.
The table sets the stated items against what each one is not (as of 2026-11-17).
| Item | What it says | Stated by | What it is not |
|---|---|---|---|
| Scope, OpenAI | reporting and archive journalism as a news source in ChatGPT, attributed short summaries and article extracts | Guardian Media Group, 2025-02-14 | not a training grant |
| Compensation | the group "explores agreements with both existing and emerging businesses to ensure fair compensation and attribution for its journalism" | Guardian Media Group, 2025-02-14 | not an amount and not a term of the OpenAI deal |
| OpenAI's aim | "access to relevant, real-time news content" | OpenAI (quote in the release) | not a value |
| Google pilot payment | pays participating publications for delivery and use of enhanced content | Google, 2026-10-15 | not a Guardian-specific term |
The compensation sentence is a statement of the group's approach to agreements in general, not a statement that a payment has been made under the OpenAI partnership. Trade coverage that uses payment phrases for the deal is not a party statement and is not used here.
The release does not mention training. That is "not stated", which is different from "no training": whether the grant reaches model training is a question for the contract text, which is not public. The same holds for the Google record.
The Guardian as a licensor and as a plaintiff
Guardian News & Media is also one of 14 named plaintiffs in Advance Local Media v. Cohere, a copyright and trademark suit filed in S.D.N.Y. on 2025-02-13 over training and retrieval-augmented outputs.
The suit is against Cohere Inc. under docket 1:25-cv-01305-CM-OTW. Its claims are the plaintiffs' allegations, and the court denied Cohere's partial motion to dismiss on 2025-11-13 (Doc. 59). The docket is in discovery as of 2026-11-02. Claims, rulings and status are tracked on the page about Advance Local Media v. Cohere.
The suit was filed one day before the OpenAI announcement. This page draws no link between the two: they involve different counterparties, and the case tracker records no OpenAI or Google litigation involving the Guardian.
A publisher can therefore license to one developer and bring a claim against another in the same period, and the records show both for the Guardian without connecting them.
What the Guardian's record shows a publisher weighing a license
The Guardian's record shows one publisher's disclosed scope, not what another publisher will be offered.
Benefits
- Attribution and extracts are named in the grant, so the way content appears in ChatGPT is stated.
- The release states an aim of fair compensation and attribution, and ChatGPT Enterprise gives the publisher a tool inside its own organization.
Risks
- Summaries and extracts sit inside a third party's product.
- The grant language leaves training unstated, so the contract text decides it.
- Enforcement against other developers runs in parallel, as the Cohere suit shows.
Because no amount is stated, a publisher cannot benchmark a price from this record. Valuation methods and the price points that are public sit on the page about what company data is worth to AI companies.
Downsides
- No amount to benchmark against.
- Pilot terms are not published per publisher.
Alternatives
- Not licensing at all.
- Display-only licensing, as in the OpenAI record.
- A per-publisher program such as Google's pilot.
- Litigation, as in the Cohere suit.
The record is also silent on the points a publisher would normally ask about, among them term, exclusivity, renewal and deletion. None of these is presented as the better route.
Publisher-wide context sits on the page about news publishers licensing content to AI companies, and the general weighing is covered under should you sell or license your data to AI companies.
Where to read the tracker, the buyers and other publisher deals
This page covers the Guardian's own deals; the pages below cover every recorded deal, the buyers and other publishers.
- AI data licensing deals with parties, dates and sources
- News archives as AI training data
- How frontier AI labs acquire data
- The Washington Post AI licensing deals
- Financial Times AI licensing deals
- AI training data market
- Google as a data buyer
Another publisher with an OpenAI record of the same kind is covered under Hearst AI licensing deals. Each page keeps its own statements with their authors, so a term stated for one publisher is never carried over to another.
Frequently asked questions
Did the Guardian license its content to OpenAI?
Yes, announced on 2025-02-14. Guardian reporting and archive journalism will be available as a news source in ChatGPT, with attributed short summaries and article extracts, according to the Guardian's release.
How much does OpenAI pay the Guardian?
No amount is stated by either party. The release gives no figure and no duration, and Google gives no per-publisher amount. Later statements are recorded in the deal tracker updates by email.
Does OpenAI train on Guardian journalism under the deal?
The release does not state training. "Not stated" is not "no training": the grant is described as availability in ChatGPT and attributed summaries, and the contract is not public.
Is the Guardian part of Google's news AI pilot?
Yes, Google's post of 2025-12-10 names The Guardian. The post gives no Guardian-specific term or value.