More than 7,000 languages are spoken in the world, but nearly half are now at risk of extinction. Since 1950, 244 have already disappeared, and within the next 40 years the rate of this loss could triple.
The disappearance of a language is as urgent a crisis as an environmental one. The mechanism is similar: pieces of cultural biodiversity are dissolving forever, with no way to recover them.
When a language disappears
Minority languages are not just different sets of words, but ways of seeing and describing the world. Over the centuries, many indigenous communities developed specific terms for plants and animals that Western science did not recognize until much later. When a language dies, all this wealth of knowledge also disappears: agricultural techniques, traditional medical remedies, rituals and concepts that simply have no equivalent in more widely spoken languages.
There is also a close link between language rights and human rights. Journalist and researcher Sophia Smith Galer, who has extensively investigated the phenomenon of “linguicide” (the deliberate erasure of a language), argues that erasing the language of a people is, in fact, affecting their identity and fundamental rights.
The digital map of inequality
Today, the historical fragility of these languages is compounded by an entirely new threat: exclusion from the digital world. A study published in October 2025 analyzed more than 7,000 languages, cross-referencing the number of people who speak them in real life with their actual presence on the Web and in technologies. The results tell us that about a quarter of the world’s languages are spoken by millions of people, but are totally absent from the data used to train artificial intelligences. The authors of the study called them “invisible giants”: languages that are alive and populated in reality, but completely excluded from the digital universe.
The avalanche effect of artificial intelligence
Artificial intelligence software (such as that behind ChatGPT) does more than just take pictures of reality: it amplifies and sometimes distorts it.
The hierarchy among languages in AI is not a simple technical error that can be solved by adding some more data. It is a structural problem. Today’s technological systems only prolong the inequalities born in colonial times, when some languages were imposed for government and trade while others were silenced.
Today, those same dominant languages fill the digital archives of the Internet, while marginal languages remain invisible to algorithms. The result is a snowball effect: the more a language is present in digital systems, the more central it becomes in society; the less it is present, the more it is cut off from those technological tools that today regulate access to information, schooling and even medical care.
Time is running out
Many local communities are struggling to save their languages through neighborhood schools, digital archives and projects to pass words from grandparents to grandchildren. But the challenge is immense and funds are scarce.
Every time an elderly minority language speaker dies, an entire world of knowledge vanishes that no computer has yet been able to save. The disappearance of languages has deep historical roots, but today it travels fast through algorithms. This is where the damage threatens to become permanent and impossible to repair.
