In an era in which artificial intelligence is “trained” daily through our use of it, a paradox emerges: the “smarter” the machine becomes, the more it reflects our weaknesses. Artificial intelligence, despite its name, is not artificial; it is the mirror of humanity’s accumulated knowledge, desires and behaviours. It is not dangerous by nature, but it can become dangerous through its use. The system accessible to the ordinary user is shaped by the questions most frequently asked, by the needs most intensely expressed, by the patterns most persistently repeated. This is what the theory of the Human–Artificial Intelligence Symbiotic Relationship (HAIST) describes as a collaborative and co-evolving relationship. Yet artificial intelligence records, filters and retransmits the social behaviours, communicative tactics and mentalities that dominate among its users. It does not learn “objectively”; it learns from us.

Conversely, artificial intelligence now shapes not only content but also perceptions; Impact Theories explain how large language models influence users’ attitudes and behaviours through the recurring patterns they have “learned” to imitate. When these patterns are simplistic or impoverished, the influence exerted becomes a reproduction of banality. In the end, does artificial intelligence produce “mirrors” that distress us when we look into them? Might the evolution we fear not be that it will independently inscribe our future in sombre letters, but rather that it may ultimately adopt our worst self?