Publishers and AI: who's signing, who's suing — and why your business should care

Fifty million dollars a year. That’s what Meta put on the table to use News Corp’s content — Wall Street Journal, New York Post, The Sun — to train its artificial intelligence. A few weeks later, nearly four hundred publishers sued OpenAI and Microsoft for doing the same thing without asking.

Same industry, same months, two opposite paths. And in between, a question that goes well beyond newspapers: how much is content worth, who sets the price — and what does the buyer actually take home?

Here’s how I read it: publishers are negotiating the price of their past, while the real stakes are the future of their relationship with people. And what’s happening to them is a dress rehearsal for anyone who produces knowledge — which today means every company.

The map: who signs, who sues

The industry is split in two, and the Reuters Institute captures it well in its 2026 journalism trends report: licensing deals are multiplying on one side, lawsuits are piling up on the other.

Among those signing: Associated Press, Axel Springer, Financial Times, Le Monde, the Guardian, Schibsted, Axios. The deals vary in nature — some cover model training, others content visibility inside answers, others access to the technology — and nearly all share one trait: the financial terms stay confidential.

Among those suing: the New York Times, which has taken OpenAI, Microsoft and Perplexity to court. Then the coalition of nearly four hundred publishers filed in New York in June. Meanwhile, Anthropic’s $1.5 billion settlement with authors set a precedent that everyone, on both sides of the table, read very carefully.

One figure, though, deflates the enthusiasm of those signing — again from the Reuters Institute: only 20% of publishers expect these deals to become a significant revenue stream. Half expect a minimal contribution. Translation: the money is real, but it’s not the game.

Italy: two philosophies under the same sky

For anyone watching from Italy, the most interesting part is that the country’s two largest publishing groups have taken opposite paths. Two completely different bets.

GEDI — the group behind la Repubblica and La Stampa — signed with OpenAI: its content, with citations and links, goes inside ChatGPT. It’s the visibility bet. The underlying reasoning seems to be: if that’s where people get their information, better to be cited inside AI answers than to disappear.

RCS — the group behind, among others, Corriere della Sera, La Gazzetta dello Sport and Io Donna — made a different choice: there is a deal with OpenAI, but it’s technological, not about content (or so it was presented, at least). Corriere doesn’t feed the engine: it uses the technology to build its own AI assistant. It’s the stewardship bet: the relationship with readers stays home.

Two groups, same city, two very different answers to the same question. And neither is trivial.

They’re not just buying words

There’s a side of these deals that gets little airtime, and to me it’s the heart of the matter.

A newspaper archive is not a warehouse of text: it’s half a century of editorial judgment. Inside it lives the knowledge of what people care about, of how to tell a story so it lands, of which language persuades and which repels, of what sells and what wins or loses elections. Whoever trains a model on that material isn’t buying words: they’re buying the craft of understanding and speaking to the public, distilled at industrial scale.

Then there’s the data flowing back — or rather, not flowing back. ChatGPT has passed 800 million weekly active users — the Reuters Institute certifies it — and the documents OpenAI filed ahead of its IPO already speak of over 900 million, with billions of prompts a day: the largest observatory ever built on what people want to know, how they ask, what convinces them. In the typical deal, the publisher hands over content and collects a fee; the platform keeps the content and all the interaction data — including how people consume that very publisher’s content. Translation: you sell the raw material and never see the seismograph.

Finally, the most uncomfortable point: whoever sits at the interface decides how and when information reaches people — which sources to select, in which words to summarise them, inside which frame, with which omissions, and at which moment. That is editorial power, exercised at a scale no newspaper editor has ever had, without necessarily the ethical and moral responsibilities we demand of an editor. The question is not whether this power will be misused. It’s: what checks exist, and who negotiates them? All the more so because international legislation, when it comes to AI, is still full of holes. So: what kind of world are we helping — or do we want to help — build?

The invisible wholesaler paradox

Here’s a point I find genuinely interesting, and it goes beyond publishing.

The more quality content you feed to answer engines, the better those engines get at answering — and the less people need to leave them. Put simply: who will buy a newspaper, a book, who will visit a website, if everything — every document, every piece of information, every answer — is already on ChatGPT, Claude, Grok and the rest? This is not a theoretical question: Google traffic to publisher sites fell by a third in a single year, and publishers expect it to nearly halve over the next three. Forbes Italia called it the risk of becoming an “invisible wholesaler”: you supply the raw material, the interface takes the relationship. Your work is there; your name isn’t.

I spent fifteen years inside multinationals with one conviction: well-made content is an investment in reputation. What I see happening to publishers confirms that conviction and radicalises it: content has become infrastructure, on a par with marketing data or — stepping outside the world of communication — energy. And as with any infrastructure, the issue is not just what they pay you for it: it’s who controls the tap and, therefore, the information.

Where I stand

I don’t think signing deals with OpenAI or the industry’s other players is wrong per se. I think signing as a wholesaler — that is, without weighing all the implications — is wrong.

And I think reducing everything to a negotiation over the fee is the gravest mistake, because the fee is the small part. The big part is control: who decides how your content reaches people — in which words, inside which frame, next to what. When you sign, you’re not just licensing an archive: you’re delegating a piece of the mediation between you and your readers to a system with its own logic, its own interests, its own opacity — one that, by the very admission of those building these technologies, nobody yet fully understands.

That’s why three clauses are worth more than the fee. Attribution: your name must stay attached to what you produce, inside the answer, not buried in a footnote. The direct relationship: no deal should intermediate it to the point of switching it off. And transparency of use: how your content is summarised and recontextualised, and which data about that use flows back to whoever produced it. A contract that pays you well today but makes you invisible tomorrow isn’t revenue: it’s a severance package.

And there’s something the sharpest publishers have already grasped, because the Reuters Institute puts it in black and white: the real answer isn’t (only) legal or commercial. It’s investing in what cannot be summarised. The investigations that lose their meaning when reduced to bullet points. The bylines people search for by name. The newsletters, the events, the communities — the places where the relationship is direct and no interface can step in between.

What this has to do with your business

Everything. Because your company produces content even if it isn’t a publisher: know-how, analysis, training, public positions. And the questions are the same.

First: your content is an asset — treat it like one. Deciding what is open, what is negotiable and what is untouchable is a strategic choice, not a legal footnote to delegate.

Second: guard your attribution. In the coming years people will ask AIs, not search engines. Being cited — with your name, your voice — is worth more than being absorbed.

Third, and most important: cultivate the channel nobody can intermediate. The direct relationship — the people who choose you, come back, trust you — is the only asset you cannot license out.

Because in the end this is the game, for newspapers and businesses alike: the ones who survive are those with an audience that loves them, not one that merely consults them. And — I say this as an author, with all the vested interest that implies — it’s true for writers too.