DataPacts

Prisa Media AI Licensing Deals (OpenAI, Meta): Terms, Value and Sources

Checked

At a glance

Prisa, the Spanish group behind El País, Cinco Días, As and El Huffpost, has three AI arrangements on record: an OpenAI agreement announced on 2024-03-13 that states model training, El País's place in Google's News AI pilot named on 2025-12-10, and PRISA's place on Meta AI's list of news content sources by 2026-07-27. Each is described by a different party.

Independent reference. Every buyer is covered on the same terms, every figure carries its source and check date, and no buyer pays to be listed or placed higher.

Request an independent introduction

Tell us what data your company holds. We reply with the buyers that fit and what their terms leave out. Your details reach a buyer only after you agree to a named introduction.

IndependentYour details stay with us until you agreeNot selling is a valid outcome

No party has disclosed an amount for any of the three. This page sets out what each party states, with its date, within the wider picture of AI training data licensing: only the OpenAI agreement states training, while Google describes AI features in Google News and Meta describes information and links.

Fact-check note: each term comes from the party document named in the sources table below. Google's and Meta's posts and El País's article were re-read on 2026-11-17; the approach is described in how we source, date and check every figure.

Prisa's three AI agreements at a glance

The three arrangements differ in buyer, date and stated use, and none carries a disclosed value.

The table lists one row per arrangement in date order, with the party that stated each term (as of 2026-11-17).

Date Buyer Prisa party as named Content Stated use Value as stated Stated by Source
2024-03-13 OpenAI Prisa Media Spanish news content: El País, Cinco Días, As, El Huffpost content in ChatGPT; "la formación y mejora continua de los modelos" not disclosed El País (Prisa Media title) elpais.com
2025-12-10 Google (Google LLC) El País enhanced news content for AI features in Google News AI-powered article overviews and audio briefings with attribution and links; training not mentioned not disclosed; Google states it pays participating publications for delivery and use of enhanced content Google blog.google
2026-07 (listed by 2026-07-27; signing date not disclosed) Meta (Meta Platforms, Inc.) PRISA real-time news information and links in Meta AI answers; training not mentioned not disclosed Meta about.fb.com

One arrangement is described by El País, a Prisa title, one by Google and one by Meta, so each statement stays with its author on this page. The parent tracker codes the rows as training stated (T) and display (D, D).

PRISA, Prisa Media and El País: which name each party uses

The group's legal notice gives the group company as Promotora de Informaciones, S.A. (PRISA), CIF A-28/297059, registered in Madrid, and the parent of Prisa Media.

The table gives the identity fields and the name each counterparty uses (as of 2026-11-17).

Field Value Source
Group company Promotora de Informaciones, S.A. (PRISA), parent of Prisa Media prisa.com legal notice (checked 2026-11-09)
Tax and register CIF A-28/297059; Registro Mercantil de Madrid prisa.com legal notice (checked 2026-11-09)
Country Spain prisa.com
Titles named in the OpenAI announcement El País, Cinco Días, As, El Huffpost elpais.com (checked 2026-11-17)
Name each counterparty uses El País: "Prisa Media"; Google: "El País"; Meta: "PRISA" elpais.com; blog.google; about.fb.com

The register number of Prisa Media as a separate company is not confirmed on this page.

El País writes "Prisa Media", Google lists "El País" and Meta lists "PRISA", so the page uses the name each party uses. It does not state which group company signed each agreement.

The OpenAI agreement of 2024-03-13: Spanish titles in ChatGPT and model training

OpenAI's agreement with Prisa Media was announced on 2024-03-13, and the announcement describes both ChatGPT and model training.

What El País reports the OpenAI agreement covers

El País, a Prisa Media title, reported on 2024-03-13 that OpenAI's agreement with Prisa Media brings content from El País, Cinco Días, As and El Huffpost to ChatGPT, and that the content will also contribute to "la formación y mejora continua de los modelos", the training and continuous improvement of the models. The agreement was announced together with Le Monde.

This is the party's own title reporting, and it is the primary source on record. The article was re-read on 2026-11-17.

No amount and no term in the announcement

Neither El País nor any other party on record states what OpenAI pays Prisa Media, how long the agreement runs or which OpenAI company signed it.

Amounts attributed to OpenAI offers appear only in press coverage and are not used here. OpenAI's other recorded agreements and their scopes sit on the profile of OpenAI as a data buyer.

El País in Google's News AI pilot since 2025-12-10

Google's post of 2025-12-10 names El País among the news publishers in a new commercial partnership pilot "to explore how AI can help drive more engaged audiences".

Google describes AI-powered article overviews on participating publications' Google News pages and audio briefings, and says the features include clear attribution and a link to articles (re-read 2026-11-17).

Google also states that it pays participating news publications for the delivery and use of enhanced content, with no amount for any participant. Its post says nothing about training.

The pilot had more than 400 participating publications as of 2026-10-15, according to Google. That scale and the other participants belong to the page on the Google News AI pilot program.

Meta's newsroom post, as updated on 2026-07-27 "to reflect the addition of new content partners", lists PRISA among the outlets it works with for real-time content in Meta AI.

The first wave of 2025-12-05 named five publications and did not include PRISA. Meta says its integrations link out to articles "while providing value to partners, enabling them to reach new audiences".

The signing date is not disclosed, so the tracker dates the entry 2026-07. Meta states no amount, no term and no training use. The full list of 13 names and their dates belongs to the page on Meta AI news licensing partners.

One group, three stated uses: how Prisa's agreements compare

Only the OpenAI agreement states training; Google describes AI features in Google News, and Meta describes information and links, without mentioning training.

The table compares the stated scope of each arrangement and who stated it (as of 2026-11-17).

Arrangement Training Display, answers or features Tracker code Source
OpenAI, 2024-03-13 stated: training and continuous improvement of the models content in ChatGPT T El País
Google News AI pilot, 2025-12-10 not mentioned article overviews and audio briefings in Google News D Google
Meta AI, listed by 2026-07-27 not mentioned information and links in Meta AI answers D Meta

"Not mentioned" is not "excluded", so the Google and Meta rows are never read as a promise that either company does not train. How publishers in general choose between training rights, display rights or both is covered on the page about news publishers licensing content to AI companies.

Which documents state each Prisa term

Every term on this page comes from three party documents and the group's legal notice.

The table compares each document and what it states (as of 2026-11-17).

Document Issuer Date What it states Last checked
"OpenAI anuncia un acuerdo con Prisa Media y 'Le Monde'" El País (Prisa Media title) 2024-03-13 parties, titles, ChatGPT, training and model improvement; no amount 2026-11-17
"Supporting the web with new features and partnerships" Google 2025-12-10 El País named in the pilot; overviews, audio briefings, attribution and links 2026-11-17
Google News AI pilot update Google 2026-10-15 payment for delivery and use of enhanced content; more than 400 publications 2026-11-17
"Bringing More Real-Time News and Content to Meta AI" Meta 2025-12-05, updated 2026-07-27 PRISA on the updated list; information and links; no money or training terms 2026-11-17
Legal notice Promotora de Informaciones, S.A. n/a company name, CIF, register 2026-11-09
OpenAI or Prisa group publication on these agreements OpenAI, Prisa none on record n/a n/a

Accounts that appear only in press coverage are not used.

El País's article was re-read on 2026-11-17 and still carries the training wording quoted above. OpenAI's own post and any Prisa release on the agreements are not on record, and personal posts are not used.

What no party to the Prisa agreements has stated

As of 2026-11-17, no party has stated the amount, the duration or the end date of any of Prisa's three AI arrangements.

  • The amount of each arrangement: El País, OpenAI, Google and Meta state none.
  • The duration and any end date.
  • Which OpenAI entity and which Prisa company signed.
  • Any OpenAI statement on its own Prisa agreement.
  • Any training statement by Google or Meta about Prisa content.
  • Any Prisa group release on the three arrangements.

A statement by a party or a passage in a Prisa annual report would close each gap.

Prisa beside Le Monde and other OpenAI news licenses on record

OpenAI's agreement with Prisa Media was announced on the same day as its agreement with Le Monde, 2024-03-13, and both sellers' accounts state that content goes into model training.

Prisa Media is one of seven seller entries in the tracker whose OpenAI license states training or model improvement, with Axel Springer, Dotdash Meredith, the Financial Times, Le Monde, Shutterstock and Vox Media, as listed on the OpenAI profile.

Le Monde's three agreements, including the Perplexity one that excludes training, belong to the page on Le Monde AI licensing deals.

What the Prisa record shows a publisher licensing training, display and pilot features

Prisa's record shows one news group licensing training use to one buyer while joining two display and feature programs, not what every publisher can obtain.

Benefits

  • A buyer that states it pays for the delivery and use of enhanced content, as Google says of its pilot.
  • Links out to the publisher's sites, as Meta describes its integrations.

Risks

  • Training use granted to one buyer, in the words of the publisher's own title.
  • A listing whose public post states no money, term or training terms.

Downsides

  • No public price to compare across the three arrangements.
  • No stated end date for any of them.

Alternatives

  • A display-only or feature arrangement with no training stated, as Google's and Meta's descriptions read.
  • An answers-only license with training excluded in writing, as Le Monde states for Perplexity.
  • Not licensing at all.

The general weighing of these options sits under whether to sell or license data to AI companies.

Where to read the buyers' records, Spanish rules and the deal tracker

This page covers Prisa's three arrangements; the tracker of every deal, Spanish rules for a training license, the buyers' records and other news deals sit on the pages below. Spanish opt-out rules, data protection and AI rules are covered under Spain's rules for licensing data for AI training.

Frequently asked questions

Does OpenAI train its models on El País content?

Yes, in the words of El País, a Prisa title. Its report of 2024-03-13 says the content will also contribute to the training and continuous improvement of the models. OpenAI has published no statement on the agreement that is on record.

How much does OpenAI pay Prisa Media?

No amount is disclosed by any party. El País, OpenAI and Prisa state no figure and no duration. New statements appear in the deal tracker updates by email.

Is El País part of Google's News AI pilot?

Yes. Google's post of 2025-12-10 names El País, and Google says it pays participating publications for the delivery and use of enhanced content, with no amount.

Does Meta use PRISA content to train Meta AI?

Meta's post does not say. It describes information and links in Meta AI answers for PRISA, listed by 2026-07-27, and does not mention training.