Taiwan Pushes Sovereign AI as a Matter of National Security
Taiwan is advancing “sovereign AI” to reduce reliance on foreign AI models. The government says it has become a national security priority, shifting to train AI models using local data. The English language currently dominates the training of AI models, making up over 50% of training data, while Chinese accounts for only 5% and is predominantly the simplified Chinese used in China, leading to fears of information bias where Taiwan is concerned.
REPORTER:
Sovereign AI is no longer just a buzzword. Taiwan is now treating it as a national security priority. The country is pushing its own homegrown AI system amid the global boom in artificial intelligence. Officials say it is key to maintaining control over data and infrastructure.
REPORTER:
Taiwan wants to reduce reliance on foreign AI models. That begins with the data used to train these models. The English language dominates, making up over 50% of training data, while Chinese accounts for only 5% -- and this is predominantly the simplified Chinese used in China, the PRC. Some analysts say this can create risks such as external influence or bias.
REPORTER:
In response to these concerns, Taiwan’s National Science Council is building a large language model that’s tailored to Taiwanese society. This involves training AI models using local data. In the future, the council hopes to secure money to help media outlets digitize historical records and donate them for training purposes.
REPORTER:
As AI development speeds up worldwide, Taiwan’s challenge is not just building artificial intelligence but deciding who controls it—and what information it is built on. The next phase will test how far Taiwan can go in building its own AI ecosystem.
