TAIPEI (Taiwan News) — The Ministry of Digital Affairs launched a campaign Tuesday to collect content that could help AI models better understand Taiwan’s values.
MODA Minister Lin Yi-ching (林宜敬) said the ministry’s corpus had accumulated about 5,000 datasets and more than 2.2 billion tokens since going online in December, per CNA. The campaign initially targets publishers and e-book platforms, with several companies and writer Lan Yi-feng (藍弋丰) signing authorization agreements.
Lin said AI models’ understanding of concepts such as democracy and power was shaped by the social context represented in their training data, per UDN. He added that international large language models still relied heavily on simplified Chinese content.
Lin said copyright issues involving new technologies could take years to resolve, but AI was developing too quickly to wait. Without more local data, he argued, both international and domestic large language models could lack a Taiwan perspective.
MODA official Wang Fu-chung (王復中) said the campaign used a royalty-free licensing model, with publishers responsible for screening and uploading content. Participation is voluntary, creators can withdraw their authorization, and licensed works may only be used for AI training.
The ministry said it was also inviting publishers, cultural institutions, and other content providers to contribute original material.
