In a sunlit conference room in Taipei, a group of Taiwanese politicians, academics, and civic leaders gathered recently for a forum organized by the Democratic Progressive Party to talk about something that did not exist a few years ago in the way it does today: the quiet, invisible power of artificial intelligence to shape what voters see, believe, and ultimately decide. The meeting was part of the party’s effort to prepare for upcoming local elections, but the tone was less about campaign strategy and more about digital self-defense. As the Taipei Times reported, participants urged the government to build databases and filtering systems specifically designed to detect and manage AI-generated content, and—just as importantly—to strengthen public media literacy so ordinary citizens can recognize manipulation before it works. There was a sense of urgency in the room, a feeling that the old rules of political communication no longer apply. In the past, voters had to worry about biased news or misleading ads. Now, with generative AI, anyone can produce realistic fake images, fake audio, and fake stories at almost no cost and at enormous scale. The question on everyone’s mind was not whether this technology would be used in Taiwan’s elections, but how deeply it already had been, and what could be done to stop it.
One of the most striking remarks came from Democratic Progressive Party Secretary-General Hsu Kuo-yung, who described a pattern he said his party had observed coming from China: information that is roughly thirty percent true and seventy percent false. That mix, he argued, is far more dangerous than a simple lie. A completely fake story can be debunked quickly, but a story wrapped around a kernel of truth is harder to untangle. It slips past fact-checkers, it feels plausible to readers, and it plants suspicion even after it has been corrected. Hsu called on experts to examine how Beijing uses artificial intelligence and big data to influence public perceptions, and he made clear that this was not an abstract problem. He was speaking about the mechanics of influence: algorithms that amplify divisive content, fake accounts that amplify certain hashtags, and AI tools that generate endless variations of the same talking point until it begins to feel like common sense. He asked the room to imagine how many hours of human labor it would take to produce millions of messages tailored to different communities in Taiwan—and then to realize that a machine could do it overnight. The forum participants did not need to be convinced. They had seen the effects in past elections: sudden spikes in angry posts, coordinated attacks on certain candidates, and rumors that spread faster than any official response could counter.
Yeh Yao-yuan, chair of the Department of International Studies at the University of St. Thomas in Houston, offered a broader strategic analysis. He said that China’s short-term goal is to influence election results—to help candidates who are seen as more favorable to Beijing, and to weaken those who are not. But the long-term goal, he warned, is even more corrosive: to deepen social divisions and undermine public trust in democracy itself. If voters come to believe that every election is rigged, that every news story is fake, and that every politician is corrupt, then the democratic system stops functioning even without military intervention. Yeh pointed to an alarming example from the United States: in July, the US government uncovered an alleged Chinese theft of 220 million US voter registration records. That data, he said, could be analyzed to identify which topics would resonate with voters in different regions—where immigration was a sensitive issue, where economic anxiety ran high, where historical grievances could be exploited. Taiwan, he argued, must assume that similar data collection is underway, perhaps on a smaller scale but with the same purpose. He called for stronger protection mechanisms, not just for government databases but for the personal information of voters, political activists, and journalists. Without such safeguards, he said, even the most sophisticated media-literacy campaigns would be fighting an uphill battle, because the AI systems would already know exactly which buttons to push.
The role of artificial intelligence in producing election interference was explored in more detail by Liu Yu-hsi of the Department of Communications at Shih Hsin University. She said that Beijing is increasingly using AI to mass-produce texts and images that can be used to interfere in elections. To illustrate, she cited online campaign materials associated with several Chinese immigrants who were registered to run for local office. At first glance, the materials looked professional and legitimate—but on closer inspection, they contained many errors. Hands had extra fingers. Backgrounds were blurred in strange ways. Text was slightly garbled in a way that a native speaker would never write. These were signs that the images and texts had been generated by AI rather than created by a human campaign team. Liu’s point was subtle and important: the errors were not just mistakes. They were evidence of a new kind of political warfare, one in which a single operator could create dozens of fake candidates, fake campaigns, and fake controversies in a matter of hours. The goal might not be to elect anyone. It might simply be to clutter the information space, to make voters weary, to create confusion about who was real and who was not, and to force legitimate candidates to waste time responding to absurd accusations. Liu said that Taiwan’s media literacy programs need to teach people to look for these telltale signs, but she also acknowledged that AI is improving quickly. What looks like a glitch today may be indistinguishable from reality in a few years. That is why she called for automated detection systems and databases of known AI-generated content, so that fact-checkers and election officials can have the tools they need to keep pace.
Arthur Wang, secretary-general of the Asia-Pacific Elite Exchange Association, offered a historical perspective on China’s tactics. He said that Beijing’s main goal has always been to undermine Taiwan’s sovereignty, but the methods have changed over time. In the past, he said, Beijing relied on verbal attacks, military threats, and economic incentives to pressure Taiwan and shape public opinion. From 2016 to 2018, the focus shifted toward disinformation—spreading fake news stories, manipulating social media trends, and funding sympathetic voices online. Now, according to Wang, the tactic has evolved again, toward targeted election interference. He described this as a more sophisticated and more dangerous strategy. Instead of trying to convince voters to support a particular candidate, the goal is to deepen divisions and hostility in Taiwanese society—to make pro-independence and pro-unification forces hate each other more, to make urban and rural voters distrust each other, to make young and old people blame each other for the country’s problems. In that kind of toxic environment, consensus becomes impossible, governance becomes harder, and the democratic process itself appears chaotic and ineffective. Wang’s remarks were a reminder that disinformation is not a single lie that can be debunked. It is a campaign to poison relationships between citizens. He urged the government and civil society to think of media literacy not just as a class in schools but as a kind of social vaccine—something that needs to be delivered to every generation, every community, and every political party, especially before elections.
Perhaps the most practical advice came from Wu Chun-chih, director of the Democratic Progressive Party’s China Affairs Department. He did not try to explain the big picture or the strategic goals of Beijing. Instead, he gave the kind of guidance that ordinary voters can use immediately. He advised treating messages from unknown sources with suspicion, especially those designed to provoke an emotional response. If a post makes you angry, or fearful, or disgusted, he said, take a breath before sharing it. Ask yourself who benefits from your reaction. If the message is designed to make you hate your neighbor or your government, it might be working exactly as intended. He recommended checking sources and verifying facts against reputable media outlets, rather than relying on forwarded messages or anonymous social media accounts. This advice sounds simple, but in the chaos of a modern election campaign, it is easy to forget. A fake story can reach millions of people before the truth has even published its first correction. Wu’s message was essentially human: be curious, be skeptical, and be willing to change your mind when new information appears. The forum ended with a shared understanding that Taiwan cannot rely on any single tool to protect its elections. Databases, filtering systems, and fact-checking networks are all necessary, but they are not sufficient. The ultimate defense is a public that understands how information works, who produces it, and why. In that sense, the challenge of AI disinformation is not just a technological problem. It is a test of civic character. In the end, the participants were asking for something profoundly democratic: the right to know what is real, the freedom to make decisions based on facts, and the power to resist manipulation without losing trust in one another.