Language, Power and Digital Identity
Artificial Intelligence has the potential to rewrite the processes of civilization: which processes and for whom, however, is not yet known.
Every civilization, in every historical era, has expressed its power through a linguistic form.
Language is never neutral: it is the primary infrastructure of sovereignty, the code with which a people thinks, names, interprets, and therefore controls reality.
When language is colonized, so is thought.
In the age of artificial intelligence, this ancient truth takes on a new and dramatic significance. Today, human language is not only a means of communication, but also the raw material of the cognitive economy: it is what generates value, information, command, and prediction.
Large Language Models (LLMs) represent the most recent and profound transformation of this relationship between language and power.
We now entrust these systems with an increasing part of our ability to understand the world. They do not merely translate or summarize: they shape reality, select what is relevant, establish connections, and order priorities. They are humanity’s new epistemological filters. Every language model is a political and ethical construction of knowledge. Whoever controls the language of machines also controls the mental form of those who use them.
Italy, a nation of culture and language, cannot and must not remain a spectator of this cognitive revolution.
It is not a question of chasing technological innovation, but of preserving our intellectual autonomy and linguistic identity, which are now threatened by a digital ecosystem dominated by models built elsewhere, according to values, categories, and priorities that are not always compatible with our history and our way of thinking.
A large-scale linguistic model is not simply software; it is, in fact, an artificial cognitive organism, trained on trillions of words, capable of learning the statistical relationships between concepts and reproducing human language in probabilistic form.
The apparent neutrality of artificial intelligence is, in reality, a political construction of knowledge in the Weberian sense.
It does not describe reality, but constructs a certain vision of reality, in line with those who designed and financed it. Every training corpus reflects a dominant culture: what is included or excluded, the weight given to one type of source over another, the interpretation underlying the concepts.
Those who use these systems implicitly accept the cognitive and moral assumptions on which they are based. A model is not just an intermediary, but a cognitive actor that filters, weighs, and guides.
This awareness has immediate implications, both culturally and in terms of national security.
The adoption of foreign linguistic models by states and public institutions entails profound risks: semantic dependence, cultural distortion, strategic vulnerability.
Already today, many Italian public offices use foreign AI platforms to draft texts, translate documents, or analyze data.
Each interaction feeds systems owned by foreign private entities, which absorb the content and reinvest it in the form of algorithmic knowledge.
This means that part of our national linguistic production, and therefore of our thinking, is exported and recycled elsewhere, losing control, context, and purpose.
It is a form of cognitive extraction: just as industrial colonialism extracted material resources from the colonies, today words, concepts, and data are extracted from cultures to feed global information economies that do not respond to national interests.
Italy, more than any other country in Europe, was born from a linguistic act. Dante Alighieri, in De vulgari eloquentia, intuited that a common linguistic community could precede a political community. When he wrote that language “is an instrument of thought,” he anticipated by centuries the modern idea that collective identity is a semantic fact even before it is a geographical one.
The political unification of 1861 was possible because, in previous centuries, Italian literary, legal, and artistic culture had already built a shared language.
Our identity was established through words. Language is a strategic infrastructure, on a par with energy or defense.
If language is the stuff of thought, then a state that does not control its linguistic tools no longer controls its collective mind.
There is also a national security dimension. Language models absorb, process, and store enormous amounts of data: legal texts, administrative documents, institutional communications. If this data is processed by foreign models, the risk of leakage, manipulation, or misuse becomes real.
Strategically, a national model would ensure data control, algorithm transparency, and cognitive resilience. It could be used for crucial applications in all ministries and especially in education to support language and digital training based on the values of our culture.
A national language model is not just an extra: it is an infrastructure of civilization that nourishes the intellectual life of a nation and expresses an act of historical continuity.
Just as Dante’s language unified Italy spiritually even before it did so politically, so an Italian language model can unite the country in the digital transition, strengthening cohesion between institutions, culture, and citizens.
It is a challenge that concerns our ability to exist in the digital world as a sovereign entity, not as a semantic periphery of either the Anglosphere or the Chinese algorithmic sphere, both of which are now hegemonic in shaping the imagery and language of globalized cyberspace.
In the field of artificial intelligence, Italy occupies an intermediate position: scientific and industrial excellence coexists with a structural lag in investment capacity and strategic vision.
For this reason, we need a state vision that recognizes digital linguistic sovereignty not as an abstraction, but as a pillar of national security and republican identity.
This article was first published on PRPChannel on November 21, 2025 and is republished with the author’s permission.