In June 2026, The New York Times and Le Monde are advocating diametrically opposed strategies toward AI platforms: one is taking legal action, while the other is redistributing 25% of its OpenAI revenue to its editorial staff. This disagreement between two globally renowned publications is not a debate about journalism. It is the clearest indication yet of what lies ahead for any industry whose value depends on quality content—including luxury brands.
By 2026, 75% of journalism jobs in the United States will have disappeared over a twenty-year period. This figure, cited by Arthur Gregg Sulzberger during a conference in Marseille, is not an alarmist extrapolation: it is the documented state of an industry that has allowed others to dictate the terms of access to its own content. The press has endured two successive shocks. It now faces a third—and this time, the players involved are two years ahead of everyone else in identifying what is happening.
The first shock was Google. In exchange for having their articles indexed, media outlets gained visibility and traffic. The market was unbalanced but understandable. The second major shift was Facebook—distribution versus visibility, which led to the collapse of local advertising and the funding of the regional press.
The third challenge is structurally different. Large language models consume journalistic content to train their systems and generate summaries—without the press receiving, in return, any traffic, revenue, or even consistent attribution. Alex Hardiman, head of products at The New York Times, puts it bluntly: “The bargain’s over.” The Web 2.0 deal is over. What applied to search and social media no longer applies. And the rules of the new regime have yet to be written.
— Details —
In 2026, a German court issued a ruling requiring Google and other operators to be held accountable for the accuracy of their AI summaries—just as a responsible publisher is held accountable for its content. If this precedent spreads across Europe, the issue of the factual accuracy of AI Overviews will cease to be an editorial matter and become legally binding. Platforms would then have an interest in sourcing and verifying their content—which means working with, and paying, producers of high-quality content.
In the face of this shift, two models coexist today among media outlets that have the resources to choose.
The New York Times has opted for a legal showdown. The organization has filed lawsuits against Microsoft and OpenAI on one front, and against Perplexity on the other. The first lawsuit has been ongoing for three years. It could take several more years to reach a resolution. The position is clear: this is a dispute over intellectual property and economics, not innovation. “Theft is not innovation,” Hardiman states categorically. The Times uses AI internally, adhering to certain principles: a journalist is always in the loop, and the technology always serves the newspaper’s mission. But it refuses to allow its content to be consumed by third parties without fair compensation. It has signed an agreement with Amazon for Alexa Plus—proof that the principle is not hostility toward platforms, but the demand for a measurable exchange of value.
Le Monde has come to a different conclusion. Louis Dreyfus, its CEO, is unequivocal about the reasons: “I don’t have the resources that The New York Times has to spend tens of millions of dollars on legal proceedings.” The economic constraint is real. But the strategy is not capitulation—it is active negotiation. Le Monde has reached agreements with OpenAI, Perplexity, and Meta, with two more expected before the end of summer 2026. These agreements define what can and cannot be used: photographs are excluded, videos are excluded, and AFP content redistributed by Le Monde is excluded so that the agency can negotiate its own rights. Whenever a model cites Le Monde, it credits Le Monde and links to its content.
What sets this model apart is not just the agreement—it’s what *Le Monde* has done with it internally. Twenty-five percent of the revenue generated by the OpenAI partnership is paid directly to the editorial staff. Every Le Monde journalist receives an annual bonus tied to these revenue streams. The decision is both economic and symbolic: it signifies that the commercial value of content is recognized as residing in the human labor that produces it, and that it is redistributed accordingly. For a publication whose editorial independence is its greatest asset, this is a governance structure, not just an accounting entry.
French regulations add another layer to the issue. At the time of this debate, Google’s AI Mode was inaccessible from mobile devices in France. The reason cited by a platform executive to Dreyfus: France is classified under the same regulatory regime as countries with which major platforms do not negotiate freely, due to the requirement to pay copyright fees before any AI synthesis can be deployed. This is not just an anecdote: it demonstrates that regulation can create an incentive to negotiate in situations where a disparity in size would otherwise make any discussion unequal.
For luxury brands, the question is no longer whether this debate concerns them. It already does.
Their photographic archives—decades of fashion shows, campaigns, and workshop portraits—are included in the training datasets of the major models. Their product descriptions, collection press releases, interviews with creative directors, and content published on third-party platforms: all of this has been indexed, ingested, and averaged. A fashion house’s proprietary terminology—industry jargon, technical terms, and internal references to a collection—now feeds models who use it without attribution, without compensation, and often without the precision needed to preserve its meaning.
It took the press ten years to clearly articulate this problem. It wasted time because it initially believed that distribution offset extraction. The lesson from 2026 is that this belief was false—and that it cost the United States 75 percent of its manufacturing jobs.
The publishing houses have an advantage that the press did not: early warning. The debate between *Le Monde* and *The New York Times* is public, well-documented, and backed by figures. The positions are clear. The economic consequences are beginning to take shape. There are now two proven models—active negotiation with internal redistribution, and a legal showdown—and their respective outcomes will become clear over the next two years.
What the publishing houses haven’t done yet is make a choice. Most haven’t even framed the question. They publish, they distribute, and they collaborate with platforms on visibility campaigns—without having defined what their content is worth under the new system, or under what conditions they can feed it into third-party models.
Dreyfus had a straightforward way of putting it: between those who have reached an agreement and those who are pursuing legal action, “there are also publishers who have neither a deal nor a lawsuit.” Those publishers, he said, are neither liable nor protected. They wait for the issue to arise on its own.
In the media, this approach has a well-documented history. In other industries with a high proportion of intangible content, that history has yet to be written.
Cette publication est également disponible en :
