The generative algorithm follows the same trajectory: its remarkable potential is likely to make it more of an infrastructural good than an optional tool – and in many cases, it already is. We can debate at length about the opacity of datasets or the monopolistic, proprietary systems, but meanwhile, universities, law firms, businesses, newspapers, and individual users are integrating its functionalities into their everyday workflows. Those who imagine boycotting AI because it was “born wrong” will soon find themselves in the position of a service worker refusing to use the internet.

In times that now seem distant, Gramsci reminded us that technology is never a neutral domain separate from culture. In Notebook 12, he writes that “there is no human activity from which one can exclude intellectual intervention; one cannot separate homo faber from homo sapiens”: if intellectual reflection is involved in every productive act, then technical literacy becomes a condition for any effective political action. In the pages of L’Ordine Nuovo, Gramsci wrote: “Industrial innovations […] allow the worker greater autonomy, place him in a superior industrial position […] The working class gathers around the machines, creates its representative institutions as a function of labour, as a function of the autonomy it has achieved, of its newly acquired consciousness of self-government” (14 February 1920). In other words, the machine becomes a factor of emancipation only when workers collectively gain control over it. In the same newspaper, a few months earlier, he insisted: “By imposing workers’ control over industry, the proletariat […] will prevent industry and the banks from exploiting the peasants and subjecting them like slaves to the safes” (3 January 1920). Technology, then, becomes a terrain of struggle: by mastering it, organised labour removes productive power from rent extraction and redirects it toward social ends. It is therefore not the machine itself that determines subordination, but the system that sets its tempo and captures its profits. From this emerges the figure of the organic intellectual: not a detached cleric, but a technical-organiser capable of uniting operational expertise with a collective project. In this light, artificial intelligence is not a gadget to be idolised or rejected, but the very ground on which a new subjectivity can test its capacity to govern productive forces and transform them toward emancipation. The urgent task is to understand and master the tool.

That is likely what AI pioneer Geoffrey Hinton meant in a recent interview when, asked what economic policies were needed to ensure AI worked for everyone, he responded with a single word: “Socialism.”

The debate on generative AI, by contrast, seems trapped in a cultural déjà-vu that Kirsten Drotner, in her studies on the history of mass media, calls media panic: an emotional cycle in which every new medium is accused of corrupting the youth, destroying employment, or undermining truth. As early as the 1990s, Drotner showed that panic follows a recurring choreography: the identification of a “vulnerable group”, a proliferation of alarmist headlines, calls for drastic measures, and finally, the medium’s assimilation into daily life. The same script has repeated itself from the 19th-century popular novel to comic books, from television to video games. Today, simply mentioning “deepfakes” or the label “AI-slop” can collapse the entire conversation into fears of an imminent cognitive catastrophe.

Yet, as Goode and Ben-Yehuda also observed, panic functions more as a ritual of symbolic stabilisation than a real risk analysis: it serves to sweep power relations off the table, shifting attention from the ownership of platforms to the morality of users. In the crosshairs end up image generators that are “too easy”, songs composed “without talent”, articles “written by a robot” – but rarely the patent chains, monopolies, or systemic exploitation that make such automation possible in the first place. The result is a short circuit: the more technology is demonised, the less we discuss governance and redistribution, leaving structural issues untouched.

Recognising media panic does not mean minimising the problems, but stripping panic of its paralysing role and returning to politics – not collective psychosis – the task of deciding who should govern artificial intelligence, and how.