Will the AI-giants end up fostering socialism?

The good thing about a paradox is that they compel us to think more deeply than usual. How could a revolution like that of AI, rooted in capitalism, lead to a world that moves away from freedom? The answer, at least a shot at it, is a curious path I want to take you through.

A few weeks ago, the German magazine Der Spiegel, which leans somewhat to the left (in the European sense), ran a cover asking whether Marx had finally been right. Will capitalism collapse, as Marx predicted, if AI is not reined in? Some of us have serious doubts about Marx as an economist and a manager of resources, but mocking Marx the philosopher and thinker is another matter; I would not fall for that temptation.

Now another economist with an almost unquestionable reputation, though The Economist took a few jabs at him recently, has published a paper that can be connected to what Der Spiegel is writing about. I am referring to Daron Acemoglu, the Nobel Prize–winning economist and MIT professor. The argument of his paper is that automation (for example, replacing workers with an AI system), which entails greater investment in capital and fuels inequality, could lead to revolts. The state can respond by redistributing wealth through taxation, regulating the automation driven by AI, or simply repressing people. What is striking is that repression becomes the most attractive option, leading the capitalist actor to finance and support a repressive regime.

Acemoglu and his colleagues' idea does not strike me as far-fetched, and although their argument is heavily mathematical, some of the events unfolding around us are worth noting. A few days ago, a post by Jacob Coxon of Anthropic, the company that created Claude, raised some alarms. He tweeted that both OpenAI (the creator of ChatGPT) and Anthropic “are not acting responsibly” and that they are engaged in a race toward superintelligence, “gambling with our lives.”

The war between the two companies became even more apparent when OpenAI announced in early September 2026 that it had solved one of mathematics' great puzzles: the Navier-Stokes equations. As it happens, two mathematicians (T. Buckmaster of NYU and L. Alpöge, an Anthropic employee) had spent a year developing a program to try to solve the problem using OpenAI's Codex platform. Strangely enough, according to what OpenAI published, 10,000 of its AI agents worked for 88 hours and arrived at what would be a solution of a similar kind before Buckmaster and Alpöge did. The question now being asked in the industry is whether OpenAI used the mathematicians' input rather improperly to beat them in the race.

It is difficult to believe that the people leading these companies are going to stop. The incentives are aligned against doing so. Meanwhile, the disappearance of white-collar jobs is moving from warning to reality, albeit with some important nuances. Without speculating about the catastrophic scenario currently being discussed (the destruction of humanity by 2030 or with a bit of luck 2032, we can, for now, consider another possibility: an economic crisis.

If many well-paid jobs slowly disappear, so will the ability of many people to service their debts, including mortgages. The more this happens, the more pressure the state will face, and that brings us back to Acemoglu: What will it do? Will it redistribute the astronomical sums these giants are earning, or will it give in to lobbying pressure to protect the interests of these companies? The idea of repression no longer seems so far-fetched considering the structural labor-market problem that could emerge.

Why talk about socialism, then? When we look at the painful lessons of history, it is difficult to separate the drama of repression from the tragedy of socialism. The latter is perfectly compatible with state capitalism, as Stalin demonstrated through his regime of five-year plans and terror, and perhaps as some countries are currently trying to emulate in more watered-down hybrid models.

If the race toward superintelligence continues at this pace, sweeping aside whoever happens to stand in its way, it is possible that the influence of social movements, translated into political parties, will grow stronger within the state amid all this uncertainty. This could lead to a crusade of good intentions, ultimately ending either in repression against the companies or, if those companies manage to influence the machinery of the state, in repression against citizens. There is always a middle ground, in which there are winners on both sides, and, as always, the ordinary citizen loses.

If we add an ingredient of group psychology, the issue becomes even more compelling. One thing that has been discovered in this field is that increasing uncertainty tends to strengthen an individual's attachment to cohesive, or “entitative,” groups. This creates two problems. One is that we end up exaggerating our differences with other groups, encouraging extremism. The other is that the more united those “opposing” groups become, the more negative judgments we form about them. In short, more polarization, at both the macro and micro levels.

If we then return —following, in part, what is sometimes called the pendulum law— to an era in which technology was sabotaged out of fear, the issue of repression ceases to be a fantasy. The question is who will seize control of the state to repress, and how will they do it, because repression is entirely feasible through the mask subtlety and smiles.

I only hope that, for the sake of humanity, the war between these giants does not prove Marx right; it would entail a great deal of human suffering. And there has already been more than enough of it.

Adaptation of my biweekly column at Revista Semana. Link