By This Hour Business Desk
Artificial intelligence companies are building products that increasingly sit between news publishers and their audiences, raising a business question with consequences far beyond media balance sheets: who pays for the reporting on which automated answers may depend?
In an opinion essay, Carine Fouteau, identified as president and publishing director of the French investigative outlet Mediapart, argues that the answer cannot be left to the market alone. She portrays the spread of generative AI as a threat to the economic base of public-interest journalism and, in turn, to the informed public debate that journalism is meant to sustain. Her proposed response is collective pressure through the courts, lawmakers and a wider alliance of journalists, researchers and citizens.
The argument is forceful, but it is also an intervention from a media executive rather than a neutral accounting of the technology sector. Its central claims turn on unresolved questions about training data, the use of publishers’ work in AI-generated responses, legal responsibility and the terms on which media organisations might be compensated. The available material does not settle those questions. It does, however, set out the case for why publishers see them as urgent.
The business dispute is about more than access to text
Fouteau’s core contention is that AI businesses benefit from journalism while weakening the publishers that fund it. In her account, AI-generated summaries can divert readers from original articles, reducing audience traffic and the revenue that supports reporting. The concern is not simply that a machine can repeat a fact. It is that a conversational product can meet enough of a user’s immediate need that the user never reaches the outlet that investigated, checked and contextualised the information.
That distinction matters because reporting is not presented as a stockpile of interchangeable words. The essay describes journalism as a process involving the cultivation of sources, collection of testimony, verification, cross-checking and editorial accountability. It argues that this work can uncover facts that powerful political or commercial interests would rather remain hidden. A system that produces answers by drawing on existing material, Fouteau argues, does not carry out that underlying reporting function.
For publishers, the commercial tension is therefore closely tied to the cost structure of their work. Original reporting requires people, time and, in some cases, exposure to danger or harassment. Fouteau points to threats faced by Mediapart’s editorial team after an investigation into Jordan Bardella, leader of Marine Le Pen’s far-right party, as an illustration of risks that cannot be reduced to the price of publishing text online. Her broader point is that an automated product can benefit from the output of expensive newsgathering without assuming comparable reporting costs or editorial obligations.
The essay contrasts the accountability of news organisations with what it characterises as a looser posture from AI companies. Publishers answer to readers and can face legal scrutiny over material they publish, it says. The author argues that AI companies seek to limit liability for the reliability of their conversational tools. The supplied material does not provide the companies’ responses to that assessment, nor does it establish how liability rules apply across jurisdictions. Those omissions are important because the legal and technical responsibilities of AI providers remain contested.
Training data has become a pressure point
Fouteau identifies the reliance of leading AI systems on copyrighted work as a potential source of leverage for publishers and other knowledge-producing institutions. The essay says OpenAI has stated that it would be impossible to train today’s leading models without copyrighted material. It also says a Microsoft vice-president acknowledged that higher-quality content improves the quality of model output when it is used in training.
Taken together, those assertions frame a simple commercial proposition: if high-quality material improves an AI product, the producers of that material may have grounds to demand control, payment or both. They also support the author’s warning that AI companies could ultimately damage the information ecosystem that supplies valuable material. In this view, a weakened reporting sector would leave AI systems with a poorer body of trustworthy work from which to generate answers.
There is a wider public-interest claim embedded in that economic argument. Fouteau says journalism orders information, establishes relevance and adds context, whereas AI-generated responses can flatten distinctions and reproduce errors or distortions. Her concern is that a public sphere increasingly mediated by algorithms could amplify misinformation, synthetic media and propaganda, encouraging anger and division rather than informed discussion.
These are consequential assertions, but they should not be read as evidence that every AI response is unreliable or that every use of news material causes publisher harm. The source material does not offer comparative performance data, quantify diverted readership, or distinguish among different AI products and licensing arrangements. It instead advances a normative argument: that the interests of profitable technology platforms and the interests served by independent reporting are not naturally aligned.
Litigation signals both resistance and expense
The essay points to legal action as one practical means of challenging AI companies’ use of journalistic material. It says The New York Times has pursued legal action, and attributes to its publisher, A.G. Sulzberger, a figure of $20 million in litigation spending. That sum, if accurately represented, conveys the scale of resources that a major publisher may need to commit when testing novel legal issues against well-funded technology businesses.
The reference also exposes a constraint in the author’s proposed strategy. Litigation can clarify rights, create negotiating leverage and potentially establish boundaries for commercial use. Yet it is costly, slow and inaccessible to many local, independent or specialist publications. A legal outcome in one place may not resolve questions elsewhere, particularly where copyright rules, related rights and civil liability differ. Court action may be a tool for some publishers, but it is not automatically a comprehensive business model for an industry under pressure.
Fouteau also says legal proceedings are under way in France, including through the body that collects press-related neighbouring rights. The supplied context provides no case names, claims, procedural posture or timetable. It is therefore not possible from the material provided to assess the likely effect of those proceedings, whether they cover training, AI-generated output or another aspect of content use, or whether they will produce a broader settlement framework.
Still, the French example fits the author’s larger call for action at national and European levels. She argues that publishers should press legislators as well as platforms, and that journalists should seek common cause with the research and knowledge communities. The aim is not merely to obtain money for publishers, in her presentation, but to defend conditions under which reliable information can be produced and made available to the public.
Collective bargaining power is the essay’s proposed answer
Fouteau does not offer a technical fix or a single regulatory demand. Instead, she calls for mobilisation on several fronts: legal challenges, political pressure and public persuasion about the value of reporting. Her argument assumes that dispersed institutions have greater leverage when they treat the issue as a shared problem rather than as a sequence of private negotiations between individual publishers and technology companies.
That approach has an evident business logic. A publisher acting alone may have limited leverage over a platform with vast distribution, computing resources and legal capacity. A coalition can pool legal costs, coordinate positions and frame the dispute as one affecting the public’s access to information, not only publishers’ proprietary interests. It may also give smaller outlets representation in debates otherwise dominated by the largest media groups and technology companies.
But collective action does not eliminate difficult choices. Publishers may differ over whether to license content, block access, seek damages, partner with AI companies or pursue a mixture of those paths. The essay does not specify how such differences should be reconciled. Nor does it detail a model for valuing journalistic work, deciding who should be paid, or measuring whether automated systems have substituted for readership rather than helped readers discover reporting.
Its most durable contribution may be to insist that those choices are political as well as commercial. If AI products become a primary route to information, their design and sourcing practices can shape which facts receive attention, whose reporting is visible and whether the organisations doing original work can continue to fund it. Fouteau’s position is that citizens have a stake in that outcome because democratic debate depends on reporting capable of scrutinising power.
A warning, not a settled account
The essay’s language is deliberately urgent, describing broad social, political, civil-liberties and environmental risks associated with large technology companies. Readers should distinguish those broad assessments from the narrower claims in the supplied material about copyrighted training data, litigation spending and proceedings in France. No counterarguments from OpenAI, Microsoft, The New York Times or the French rights-collection organisation are included here, and the legal claims have not been detailed enough to test independently.
The report has not been independently corroborated. It is based on a source-limited opinion essay and should be understood as Fouteau’s case for collective resistance to AI companies’ use of journalism, rather than a verified account of the underlying disputes or a prediction of their outcome.
Even so, the question posed by the essay is likely to persist: whether the economic value created by original reporting can be preserved when automated systems increasingly package information for users. The eventual answer may be shaped by courts, legislation, licensing negotiations and public preferences. The available material establishes only that publishers such as Mediapart see the issue as central to their future, and that some are seeking to contest it rather than accept the terms offered by technology platforms.
For further context on this subject, see Report says FTC examines security incidents involving Anthropic and OpenAI.
Reporting notes
What is confirmed: The supplied source says OpenAI and a Microsoft executive made comments about copyrighted and high-quality training material, and that publishers are pursuing legal avenues.
Why this matters: The dispute links copyright, AI product quality and the financial sustainability of public-interest journalism.
What remains unclear: The underlying statements, litigation details, legal outcomes and the scale of any publisher harm were not independently verified in the supplied material. This report is based on one source and has not been independently corroborated.
