How China used TikTok, AI, and big data to target Taiwan's elections

China cultivated TikTok influencers in Taiwan and elsewhere to spread specific disinformation, even as U.S. lawmakers and free-speech advocates debate threats posed by the social-media platform, Taiwanese researchers said Monday.

Beijing has used government-related accounts and generative AI to push fake videos, and used third-party and other data to micro-target messages and influence campaigns to specific individuals. The researchers said China’s efforts in Taiwan provide a glimpse into tactics and techniques they could use on a global scale to undermine support for the U.S. government and military inside and outside the country. 

Leading up to Taiwan's election in January, China did “use Taiwanese voices to discredit Taiwan, cultivate official brands or personalities or influencers” to push disinformation streams, Chihhao Yu, a software designer and co-director of the Taiwanese Information Environment Research Center, or IORG, told reporters. 

Because many TikTok influencers traffic in disinformation as a simple business practice, IORG wanted to find a stronger link. They looked at influencer-shared content frame by frame and even pixel by pixel to show that the source was likely the Chinese government. Often, groups of influencers across multiple countries might release similar content all at the same time, related to an issue the Chinese Communist Party was trying to influence. 

“That's an even stronger signal indicating that these influencers on the site are having at least some kind of coordination with PRC actors,” he said.

China relies on a variety of conspiratorial themes in their messaging, such as that the United States is attempting to build bioweapons labs on Taiwan, that it is seeking to draw China into conflict, and that the United States is a poor ally because it's on the verge of a civil war. 

Eve Chiu, editor-in-chief of the Taiwan Factcheck Center, pointed out efforts to use generative AI to create fake videos showing Taiwenese election authorities stuffing ballot boxes and even destroying ballots. She also highlighted a deepfake video that purported to show U.S. Rep. Rob Wittman, R-Va., saying that the U.S. military had endorsed particular Taiwanese candidates. 

Billion Lee, co-founder of the public disinformation monitoring group Cofacts, said China is able to use big data, third party data, and more to target specific individuals with posts and disinformation threads, but right now that’s only about 10 percent of Chinese disinformation posts, she said. 

We-Ping Liu, part of Taiwan’s Ministry of Justice investigation Bureau, said China still has to use local human “collaborators” across China and elsewhere to make specific posts on social media convincing, rather than sounding like bots.  

That suggests that wider adoption of AI, and particularly large language models, could greatly improve the effectiveness of China’s disinformation campaigns, particularly when it comes to English-speaking countries.