The launch of “DeepZang” in Lhasa is being promoted as a milestone for the Tibetan language a state-backed AI system designed to translate, write, and process Tibetan alongside Chinese and English. Officials present it as a tool of preservation and progress.
Yet beneath this narrative lies a harder truth.
Language is more than utility; it carries memory, belief, and identity. In Tibet, that foundation has already been strained. Educational policies have increasingly prioritized Mandarin, leaving many Tibetan children with limited fluency in their own mother tongue. Against this backdrop, the emergence of a state-controlled Tibetan AI model raises urgent concerns.
DeepZang does not simply process language it defines it. The data it is trained on, the content it filters, and the knowledge it presents are all shaped within a tightly controlled system. This means the model cannot be neutral. It reflects the priorities and boundaries set by those who built it.
Supporters argue that such technology brings Tibetan into the digital age. But digitization without autonomy is fragile. A language survives through daily use, community transmission, and cultural freedom not only through algorithms.
There is a clear contradiction: while Tibetan language use declines in classrooms and communities, it is simultaneously being formalized and managed in digital space. This risks turning a living tradition into a standardized, state-approved version of itself.
The question is no longer whether Tibetan can exist online. It is whether Tibetans themselves can shape how their language lives, evolves, and endures.
