In his column Face à Michel Onfray on CNews, the philosopher revisits Proudhon’s formula, “property is theft,” to examine the exploitation of labor in the age of artificial intelligence. According to his analysis, AI systems benefit from texts and research produced by authors without ensuring they are compensated, a practice that undermines creators and the entire book industry.
Pierre-Joseph Proudhon’s formula is often reduced to a general condemnation of property. Michel Onfray proposes recovering its meaning by placing it in context: the 19th-century thinker was primarily attacking the appropriation of other people’s labor and of a share of the wealth created collectively but not returned to those who produced it.
This interpretation leads the columnist to draw a parallel with artificial intelligence. Generative tools rely on vast quantities of texts, images, and other content created by people. Yet, according to Onfray, authors whose works feed these systems do not necessarily receive compensation when their output contributes to the creation of a commercial service.
A provocative formula, but a precise critique
To understand Proudhon’s thought, Onfray urges us to distinguish property as such from the appropriation of the fruits of labor. In his analysis, the author of What Is Property? was targeting a form of dispossession: the owner benefits from work done by others without paying them the value corresponding to their contribution.
The example of the obelisk in Place de la Concorde illustrates this reasoning. Its installation required the coordination of many men working toward the same goal. The collective force mobilized for an hour cannot simply be reduced to that of a single worker over an equivalent period multiplied by the number of participants. For Proudhon, the organization of shared labor produces a particular value, which can be seized if it is neither recognized nor compensated.
In this interpretation, the formula “property is theft” thus refers to the possibility that an individual may appropriate the result of a collective effort. Onfray points out that Proudhon was not, however, an advocate of abolishing all property. He was wary of collectivism and notably defended small-scale property, which he regarded as protection against the concentration of power and forms of state confiscation.
From collective labor to content used by AI
The comparison with AI rests on the question of contribution and compensation. Generative models are trained on vast corpora, made up in particular of content published online. These materials are the fruit of the work of journalists, writers, researchers, photographers, and many other creators. Once incorporated into a system, they can contribute to answers or outputs provided to users, sometimes as part of a paid service.
Onfray describes this situation as a form of appropriation of intellectual labor. The technology service derives economic value from resources produced by third parties, while the authors concerned have not always been consulted, compensated, or informed about how their works are used. The analogy with Proudhon is therefore less about a specific legal accusation than a political question: who gets the value created through a shared effort?
Debates surrounding AI now extend beyond individual uses and concern the rules governing the circulation of content. The European developments discussed in this article on regulatory changes relating to artificial intelligence and digital content illustrate the issues of transparency, rights, and responsibility that accompany the development of these tools.
Copyright at the heart of the debate
Onfray’s reflections also build on Proudhon’s writings on literary property. In Literary Majorats, the philosopher examines the rights granted to authors and the limits that should be placed on them. The principle defended in the column is that writers should be able to earn compensation for their work, without their works thereby becoming wealth that can be passed on indefinitely, detached from any collective contribution.
This position highlights a tension that remains ever-present. A work bears the mark of its author, but its creation also depends on the language, knowledge, techniques, and exchanges that make writing possible. The question is therefore not merely whether a text belongs to someone: it also involves determining how to recognize the author’s work and regulate the uses of their work, including when it is included in databases used to train models.
Works may be accessible, but not necessarily free to use
The availability of content on the Internet does not mean it can be reused without conditions. Onfray notes that some of his own books circulate free of charge on websites, sometimes without his consent. He links this unauthorized distribution to the broader problem posed by the large-scale exploitation of cultural and intellectual output.
The comparison does not mean that every situation falls under the same legal framework. Downloading a book, training an AI system, and generating text from a model raise distinct questions. But in his argument, these practices have in common the possibility of separating the use of a work from compensation for the person who created it.
The discussion also concerns how institutions and companies make the conditions governing the use of content visible. In the United States, the initial protective measures adopted in California provide an example of local regulation: California introduces safeguards around artificial intelligence. These initiatives reflect efforts to establish guardrails for a technology whose effects are unfolding rapidly.
A threat to authors and the publishing industry
In the perspective developed by Onfray, the lack of compensation affects more than individual incomes. It can affect the very conditions under which knowledge is produced. If creators see their content reused without financial return, they have fewer resources to continue their research, write, or publish. Precarity can then affect all professions involved in producing works and knowledge.
The columnist connects this development to the difficulties facing the publishing industry. Falling sales, more limited risk-taking, and the difficulty of publishing specialized works could, in his analysis, reduce the diversity of books available. Publishers need to be able to invest in books that sometimes have a limited readership but make an important intellectual contribution.
The fear expressed is that the book will gradually be weakened as a space for transmitting and developing ideas. If content can be easily summarized, recombined, or made accessible through automated tools, the economy that enables its creation could contract. Onfray fears that an abundance of generated texts will eventually occupy the space at the expense of works produced through long, documented, and identifiable labor.
The quantity of answers versus the quality of knowledge
The critique also targets the possible confusion between the accumulation of information and intelligence. AI systems can quickly produce structured answers, but this fluency alone does not guarantee the reliability of sources, the soundness of reasoning, or mastery of a subject. Onfray warns against an era in which the ability to generate text might be mistaken for proof of knowledge.
An automated model can aggregate very different kinds of content without necessarily distinguishing the work of a recognized expert from that of an author whose expertise is recent or uncertain. Users must therefore maintain a critical perspective on what they receive: check references, identify the limits of the answer, and avoid mistaking confident wording for an established argument.
This concern connects with debates about the use of AI-related vocabulary in public life. An article about Donald Trump’s statements reports, in particular, that he called those who use the term “enemies” in certain exchanges: the controversy surrounding his remarks on artificial intelligence shows how technology has also become a subject of political confrontation.
The technological revolution and the battle over value
The debate over authors’ labor is part of a broader economic competition. Companies are investing in infrastructure, chips, data centers, and models capable of processing ever-greater volumes of information. AI is therefore transforming not only the production of texts: it is also contributing to a reorganization of markets and technological power dynamics.
The stakes are reflected in the financial moves of major companies in the sector. The $40 billion financing project associated with SpaceX, presented as potentially strengthening its role in artificial intelligence, illustrates the scale of the capital being deployed: SpaceX’s proposed AI-related deal gives an indication of the ambitions driving this industry.
This concentration of investment echoes the question raised by Proudhon: who creates value and who reaps the benefits? Content creators contribute intellectual material, while companies develop the tools and services that make it possible to exploit it on a large scale. The distribution of revenue, recognition of contributions, and terms of authorization thus become central issues.
Uses expanding into other sectors
The expansion of AI also affects fields far removed from writing. Changes in the automotive industry, for example, intersect with issues of automation, electrification, and the strategies of technology companies. An article on artificial intelligence and competition in the electric car market shows that this revolution is spreading to sectors where data, software, and computing power are playing an increasingly important role.
In this context, Onfray’s critique invites us not to reduce the question to an opposition between progress and rejecting progress. Rather, it is a matter of examining the concrete conditions under which technologies are developed: what content is used, who controls its reuse, what compensation is provided, and what protections are in place for the people who produce the resources exploited by these systems.
Proudhon’s thought thus retains contemporary relevance. It allows us to question how the fruits of labor are shared when many people contribute to it, even when that labor has become digital, dispersed, and difficult to identify. In Michel Onfray’s column, the historical formula serves to examine a current transformation: the value of works can be captured by powerful tools, while their authors sometimes remain excluded from the benefits they help generate.
