Taiwan's Central News Agency (CNA) announced that an AI chatbot using public media news materials as its primary knowledge source will officially launch by the end of 2026. The project aims to help rebuild social trust amid fake news and disinformation campaigns.

Chinese Television System (CTS) Chairperson and project initiator Hu Yuan-hui (胡元輝) said Monday that Taiwan’s public media outlets are committed to building a trustworthy AI chatbot, calling it a milestone in the nation’s media history. He noted that major international outlets such as the Financial Times and Washington Post have already launched AI chatbots, inspiring him to provide the Mandarin-speaking world with a similarly reliable source of information about Taiwan and the world.

Developed in collaboration with the Taiwan Broadcasting System (TBS) and Academia Sinica, the chatbot draws from the news databases of CNA, Public Television Service (PTS), and Radio Taiwan International (Rti), along with other credible sources authorized by the government and NGOs. Hu emphasized that this distinguishes their chatbot from typical AI models, which rely on a wide range of online sources, making quality and accuracy harder to guarantee.

The launch coincided with an announcement that domestic media outlets have formed an alliance to regulate online news. Rti Chairperson Cheryl Lai (賴秀如) said this represents a new wave of self-regulation following Taiwan’s media liberalization. CNA has also introduced an AI editing assistant designed for the news industry to aid in fact-checking, typo detection, and caption recommendations.

Lai added that RTI will sign a partnership agreement with Google this week to use AI technology to develop a citizen journalism forum for foreigners in Taiwan. “Everyone is trying to find a niche in AI. Our effort is about rebuilding trust in society as a whole. If a democratic society is plagued by unrest due to its inherent freedoms, we need to leverage more expertise to foster trust. In doing so, we can further strengthen our democracy,” she said.

Academia Sinica Institute of Information Science professor and project technical advisor Ku Lun-wei (古維倫) noted that the amount of AI training data in Taiwan is comparable to that of some African languages but far less than for English. She said compiling news data from so many public media sources to train a native Taiwanese AI chatbot is a highly significant step.