
The unveiling of “DeepZang” – rendered in Tibetan as “Tér-Tsang,” meaning “Treasure House” – and widely described in Chinese media as the world’s first Tibetan large language model, marks a significant milestone in the evolution of artificial intelligence for underrepresented languages. Developed by Tenzin Norbu’s Tibet Choknor Information Technology Co., Ltd. and launched in Lhasa on 16 March 2026, the platform reflects both technological advancement and a growing recognition that the Tibetan language deserves robust digital infrastructure in the age of AI.
On its face, DeepZang is a welcome and necessary development. A system capable of processing Tibetan language across more than 80 languages – integrating speech, translation, and multimodal capabilities – has the potential to expand access to education, preserve linguistic heritage, and reinforce cultural continuity in an increasingly digitised world. At a time when smaller languages risk marginalisation in the global AI landscape, such initiatives can help ensure that Tibetan remains a living, evolving medium of knowledge and expression.
Yet this moment must also be understood in parallel with the determined and longstanding efforts of Tibetans themselves to harness technology in service of their language and culture. Long before DeepZang’s launch, Tibetan AI applications were already being developed in exile. The work of the Monlam Tibetan IT Research Centre in Dharamsala – founded by the Tibetan monk-scholar Geshe Lobsang Monlam – stands as a powerful example of this commitment. Its platforms, including Melong.ai and the upgraded “Monlam Think,” have pushed the boundaries of Tibetan-language AI, reportedly surpassing several global models in comprehension and generation while supporting over 140 languages and multimodal inputs. These initiatives are not merely technical achievements; they are conscious efforts to safeguard Tibetan language, history, and culture in the face of ongoing challenges.

At the same time, it is notable that the current Tibetan parliamentary session in Dharamsala has been actively engaging with questions of artificial intelligence in governance and the effective functioning of the Central Tibetan Administration (CTA), Tibet’s government-in-exile. Within its ongoing budgetary deliberations, discussions have included allocating resources and securing funding for AI-related initiatives. This reflects a growing recognition among Tibetan leadership that digital tools, including AI, will play a critical role in governance, public service delivery, and the long-term preservation of Tibetan civilisation.
However, the emergence of DeepZang must also be situated within a broader political and legal context. Artificial intelligence systems do not operate in isolation; they are shaped by the regulatory and ideological environments in which they are built and deployed. In China, where information governance is closely aligned with state priorities, concerns inevitably arise about how such technologies may function – particularly on issues relating to Tibetan religion, history, and identity.
These concerns are sharpened by recent legislative developments. The “Ethnic Unity and Progress Law,” passed by the National People’s Congress on 12 March 2026 and set to come into effect on 1 July 2026, reflects a policy trajectory that many view as deepening assimilation efforts. Within this context, the preservation of Tibetan language and identity faces increasing pressure, as policies continue to prioritise uniformity over cultural plurality.
A central issue, therefore, is whether data deemed sensitive by the Chinese Communist Party (CCP), when entered into the DeepZang system, will produce full and accurate outputs – or whether it will be filtered, reshaped, or blocked altogether by the country’s extensive information control architecture, often referred to as the “Great Firewall.” If such constraints are embedded within the system, the platform’s ability to function as an open and reliable repository of Tibetan knowledge may be fundamentally limited, particularly on issues that lie at the heart of Tibetan identity.
The concern, then, is not with innovation itself, but with its governance. If user queries and outputs are monitored or aligned with predetermined narratives, platforms like DeepZang risk functioning as instruments of soft control rather than open knowledge systems. This raises fundamental questions about user autonomy, data privacy, and the right to access and engage with information freely.
In contrast, the efforts of Tibetan developers in exile underscore a different vision – one in which technology serves the people, preserves authenticity, and enables open inquiry. Their work highlights how AI can be aligned with the protection and revitalisation of cultural identity, rather than its dilution or reinterpretation through external control.
Taken together, these developments illustrate both the promise and the tension at the heart of Tibetan AI today. DeepZang should be recognised as a technological achievement, but it must also be evaluated critically in light of the broader political environment in which it operates. At the same time, the contributions of Tibetan innovators – particularly those working to safeguard language, history, and culture under challenging circumstances – deserve far greater recognition and sustained support.
As artificial intelligence continues to shape the future of knowledge and communication, one principle should remain clear: technology must empower people to think, question, and preserve their identity – not constrain them. For Tibetans, this is not only a question of innovation, but of cultural survival.
Tsering Passang is a London-based Tibetan blogger and the founder–chair of the Global Alliance for Tibet & Persecuted Minorities. A long-standing Tibetan human rights advocate, he works internationally to advance justice, freedom, and peaceful solutions for Tibetans and other persecuted communities living under authoritarian rule. His writing can be found at www.Tsamtruk.com.







































































You must be logged in to post a comment.