As part of its discussions with major players in the field of artificial intelligence, Le Monde has just signed a multi-year agreement with OpenAI, the company known for its ChatGPT tool. This agreement is historic as it is the first signed between a French media organization and a major player in this nascent industry. It covers both the training of artificial intelligence models developed by the American company and answer engine services such as ChatGPT. It will benefit users of this tool by improving its relevance thanks to recent, authoritative content on a wide range of current topics, while explicitly highlighting our news organization's contribution to OpenAI's services.
This is a long-term agreement, designed as a true partnership. Under the terms of the agreement, our teams will be able to draw on OpenAI technologies to develop projects and functionalities using AI. Within the framework of this partnership, and for the duration of the agreement, the two parties will collaborate on a privileged and recurring basis. A dialogue between the teams of both parties will ensure the monitoring of products and technologies developed by OpenAI.
For the general public, the effects of this agreement will be visible on ChatGPT, which can be described, in simple terms, as an answer engine using established facts or comments expressed by a limited number of references. The engine generates the most plausible and predictive synthetic answer to a given question.
The agreement between Le Monde and OpenAI allows the latter to use Le Monde's corpus, for the duration of the agreement, as one of the major references to establish its answers and make them reliable. It provides for references to Le Monde articles to be highlighted and systematically accompanied by a logo, a hyperlink, and the titles of the articles used as references. Content supplied to us by news agencies and photographs published by Le Monde are expressly excluded.
Consolidating our business model
For Le Monde, this agreement is further recognition of the reliability of the work of our editorial teams, often considered a reference. It is also a first step toward protecting our work and our rights, at a time when we are still at the very beginning of the AI revolution, a wave predicted by many observers to be even more imposing than the digital one. We were among the very first signatories in France of the "neighboring rights" agreements, with Facebook and then Google. Here too, we had to ensure that the rights of press publishers applied to the use of Le Monde content referenced in answers generated by the services developed by OpenAI.
This point is crucial to us. We hope this agreement will set a precedent for our industry. With this first signature, it will be more difficult for other AI platforms to evade or refuse to negotiate. From this point of view, we are convinced that the agreement is beneficial for the entire profession.
Lastly, this partnership enables the Société Editrice du Monde, Le Monde's holding company, to work with OpenAI to explore advances in this technology, anticipating as far as possible any consequences, negative or favorable. It also has the advantage of consolidating our business model by providing a significant source of additional, multi-year revenue, including a share of neighboring rights. An "appropriate and equitable" portion of these rights, as defined by law, will be paid back to the newsroom.
These discussions with AI players, punctuated by this first signature, are born of our belief that, faced with the scale of the transformations that lie ahead, we need, more than ever, to remain mobile in order to avoid the perils that are taking shape and seize the opportunities for development. The dangers have already been widely identified: the plundering or counterfeiting of our content, the industrial and immediate fabrication of false information that flouts all journalistic rules, the re-routing of our audiences towards platforms likely to provide undocumented answers to every question. Simply put, the end of our uniqueness and the disappearance of an economic model based on revenues from paid distribution.
These risks, which are probably fatal for our industry, do not prevent the existence of historic opportunities: putting the computing power of artificial intelligence at the service of journalism, making it easier to work with data in a shorter timeframe as part of large-scale investigations, translating our written content into foreign languages or producing audio versions to expand our readership and disseminate our information and editorial formats to new audiences.
Protecting our editorial identity
To take the measure of these challenges, we decided to act in steps. The first was devoted to protecting our content and strengthening our procedures. Last year, we first activated an opt-out clause on our sites, following the example of several other media organizations, prohibiting AI platforms from accessing our data to train their generative intelligence models without our agreement. We also collectively discussed and drew up an appendix to our ethics and deontology charter, devoted specifically to the use of AI within our group. In particular, this text states that generative artificial intelligence cannot be used in our publications to produce editorial content ex-nihilo. Nor can it replace the editorial teams that form the core of our business and our value. Our charter does, however, authorize the use of generative AI as a tool to assist editorial production, under strictly defined conditions.
With this in mind, another phase was opened, dedicated to experimenting with artificial intelligence tools in very specific sectors of our business. Using DeepL, we were able to launch our Le Monde in English website and app, whose articles are initially translated by this AI tool, before being re-read by professional translators and then edited and published by a team of English-speaking journalists. At the same time, we signed an agreement with Microsoft to test the audio version of our articles. This feature, now available on almost all our French-language articles published in our app, opens us up to new audiences, often younger, as well as to new uses, particularly for people on the move. The third step is the one that led us to sign the agreement with OpenAI, which we hope will create a dynamic favorable to independent journalism in the new technological landscape that is taking shape.
At each of these stages, Le Monde has remained true to the spirit that has driven it since the advent of the Internet, and during the major changes in our industry: We have sought to reconcile the desire to discover new territories, while taking care to protect our editorial identity and the high standards of our content. In recent years, this approach has paid off. As the first French media organization to rely on digital subscriptions without ever having recourse to online kiosks, we have for several years been able to claim a significant lead in the hierarchy of national general-interest dailies, thanks to an unprecedented number of over 600,000 subscribers. In the same way, our determination to be a pioneer on numerous social media platforms has given us a highly visible place on all of them, helping to rejuvenate our audience.
The agreement with OpenAI is a continuation of this strategy of reasoned innovation. And we continue to guarantee the total independence of our newsroom: It goes without saying that this new agreement, like the previous ones we have signed, will in no way hinder our journalists' freedom to investigate the artificial intelligence sector in general, and OpenAI in particular. In fact, over the coming months, we will be stepping up our reporting and investigative capabilities in this key area of technological innovation.
This is the very first condition of our editorial independence, and therefore of your trust. As we move forward into the new world of artificial intelligence, we have close to our hearts an ambition that goes back to the very first day of our history, whose 80th anniversary we are celebrating this year: deserving your loyalty.