As the United Nations’ International Decade of Indigenous Languages (2022–2032) reaches its midpoint, linguists and Indigenous communities worldwide are turning to an unlikely ally to save the roughly 3,000 languages currently at risk of disappearing: artificial intelligence. Recent reports highlight a growing number of collaborations in which Indigenous speakers, computational linguists, and tech companies are co-developing language models, speech recognition tools, and digital archives — efforts that could fundamentally reshape how humanity preserves its linguistic heritage.
A Crisis Decades in the Making
According to UNESCO, a language disappears roughly every two weeks, taking with it generations of cultural knowledge, oral history, and unique ways of understanding the world. Of the world’s approximately 7,000 languages, nearly half are considered endangered. Indigenous languages are disproportionately affected, with many spoken by only a handful of elderly speakers. The reasons are well documented: colonial-era assimilation policies, urbanization, the dominance of major world languages in education and media, and economic pressures that push younger generations away from heritage tongues.
The launch of the Decade of Indigenous Languages was meant to galvanize global action, but progress has been uneven. While countries such as New Zealand have made significant strides reviving te reo Māori, others continue to lose speakers faster than documentation efforts can keep pace. That asymmetry has prompted communities to look beyond traditional fieldwork methods.
How AI Is Changing the Equation
Machine learning tools that once required millions of data points are now being adapted for “low-resource” languages — those with minimal written records or digital text. Researchers at institutions including the Endangered Languages Project, a collaborative initiative supported by Google and academic partners, have begun training models on smaller datasets, often built directly with native-speaker elders. These models can transcribe oral recordings, generate teaching materials, and even allow learners to converse with chatbots in languages that have no equivalent in mainstream platforms like Google Translate.
One notable example is the work being done with Cherokee, Māori, and several Aboriginal Australian languages, where speech-to-text systems are being fine-tuned on community-recorded audio. The First Languages AI Reality (FLAIR) initiative at the Mila AI institute in Montreal, led in part by Indigenous researcher Michael Running Wolf, has been particularly vocal about ensuring that these tools remain under community control rather than extracted by outside corporations.
Tensions Over Data Sovereignty
The push toward AI-assisted revitalization is not without controversy. Indigenous scholars have raised pointed questions about who owns language data once it is digitized, and whether large tech companies should be allowed to train commercial models on sacred or culturally restricted material. The concept of “Indigenous data sovereignty,” articulated through frameworks such as the CARE Principles for Indigenous Data Governance, has become a guiding standard for many projects, emphasizing collective benefit, authority to control, responsibility, and ethics.
Linguist Wesley Leonard of the Miami Tribe of Oklahoma has argued that revitalization is not simply a technical problem of preserving phonemes and grammar but a deeply social one — about restoring relationships between people and their ancestral knowledge. AI, in this view, is a tool, not a solution. Communities that lose oversight of their linguistic data risk seeing their heritage commodified or misrepresented, an outcome that mirrors earlier waves of cultural extraction.
Why It Matters Beyond Linguistics
The stakes extend well beyond academia. Languages encode ecological knowledge crucial for biodiversity conservation, traditional medicine, and climate adaptation. When a language dies, so often does an irreplaceable understanding of local ecosystems — a loss researchers increasingly link to the broader environmental crisis. Reviving these languages is therefore not only a cultural project but, many argue, a scientific and ecological one.
Looking ahead, the next five years of the UN Decade will likely determine whether AI becomes a genuine equalizer or another extractive technology layered atop centuries of dispossession. Watch for forthcoming UNESCO progress reports in 2026, expanded Indigenous-led AI labs, and new legal frameworks around language data — particularly in Canada, Australia, and parts of Latin America. The outcome will shape not just the survival of thousands of languages but the very question of who gets to decide how human knowledge is preserved.
For more stories at the intersection of language, culture, and science, visit science.wide-ranging.com for related coverage and in-depth reporting.
