The cyber proletarian of the 21st century will not be deprived of factories, but of augmented intelligence.
For a long time, artificial intelligence has been presented as a promise of democratization. Everyone, from isolated researchers to small entrepreneurs, could access tools of unprecedented power. Of course, you had to pay a subscription, but the border remained relatively porous. AI seemed to usher in a new era where intellectual capital would be determined less by the power of the organization than by the creativity of individuals. This illusion is dissipating.
With its Glasswing project, Anthropic has announced that it is reserving Claude Mythos for a restricted circle of clients, large tech companies, cybersecurity players and hand-picked critical infrastructure operators. Same logic at OpenAI, which already reserves certain capacities, calculation quotas or early access to its most strategic customers. Scarcity – which is also an excellent marketing technique – is no longer technical, it becomes political and economic. The best tools will not be accessible to everyone, but distributed according to the purchasing power, influence and commercial value of the customer.
We are witnessing the creation of a new aristocracy: that of the holders of cognitive power
On the one hand, a few large companies, governments and organizations with privileged access to artificial intelligence capable of reasoning faster, coding better, analyzing more deeply, simulating more and deciding with formidable efficiency. On the other, the mass of ordinary users condemned to use restricted versions, slower, less precise, less autonomous. In other words, after the means of production, here are the means of cognition.
Karl Marx described the proletarian as one who does not possess the tools of production and must sell his labor power. The cyber proletarian of the 21st century will not be deprived of factories, but of augmented intelligence. He will work with inferior tools, facing competitors with superhuman digital assistants. With equal skills, he will produce less, more slowly and with less relevance. And he will be reduced to renting out a few artificial cognitive capacities all his life, billed for use by a handful of companies controlling the global intelligence infrastructure. An infrastructure – and this is the irony of the story – which will be largely supplied by its own energy resources, captured at a low price by the new masters.
The digital divide, yesterday measured in access to the Internet, will shift towards access to computing power and cutting-edge models. Some will have invisible armies of agents capable of drafting, scheduling, negotiating, monitoring and anticipating. The others will remain alone in front of their keyboard.
The consequences will be considerable
In cybersecurity, this advantage will be decisive. These advanced artificial intelligences will make it possible, on the defense side, to massively reduce technical debt thanks to an almost exhaustive search for vulnerabilities, configuration errors and potential exploitation chains. But they will also offer, on the offensive side, new capabilities to those who choose to keep these vulnerabilities for themselves in order to exploit them. This strategic question undoubtedly explains the quarrel between Anthropic and the American Department of Defense.
But, more broadly, the consequences are first and foremost economic. The productivity of the best-equipped organizations will soar, reinforcing the concentration effects already observed in digital technology.
The consequences are then social. A new hierarchy will appear within the intellectual professions themselves, separating “augmented” workers from “bare” workers.
Finally, the consequences are also geopolitical. States mastering the most advanced infrastructures, semiconductors and models will impose their standards, their dependencies and, ultimately, their vision of the world.
Some had promised us an emancipatory revolution. We see the emergence of a society of cognitive castes.
Of course, the story is not completely written. Open source models, European initiatives and ambitious public policies can still prevent the extreme concentration of this power.
But it is clear, as Arthur Mensch, co-founder and CEO of Mistral AI, explained during his hearing at the National Assembly, that Europe has fallen behind in this gigantic race to capture value. It must therefore activate without delay all available levers — including by reopening the debate on pension funds — while ceasing to consider regulation as the alpha and omega of an industrial policy worthy of the name.
Cynically, we can also console ourselves by considering that this massive delegation of our faculties to artificial systems will generate in those who abuse it a profound dependence, a progressive loss of essential skills – because, before being “augmented”, we still need to know how to create and act “naked” – and, ultimately, a form of cognitive and physiological atrophy. Without forgetting a skills “pipeline” problem in a few years.
But we must face reality: in the short and medium term, artificial intelligence will only accentuate the concentration and accumulation of economic, intellectual and political power in the hands of the United States and China. And will massively strengthen the capabilities of augmented organizations and workers. Moreover, as the saying attributed to Pablo Picasso says, “artists create, geniuses plunder”…
The question is no longer who owns the machines, but who owns the artificial minds. And when access to intelligence becomes a privilege, equality of opportunity is a distant memory.
Cyber proletarians of all countries, unite! You have nothing to lose… except a few tokens.
