Insider》23 million users in Taiwan have been profiled, China divides into "four major groups" to launch AI united front
China systematically collects 23 million Taiwanese household registration and account information, and uses AI big data to accurately classify four major groups and deliver customized messages.
Quick Look
- National security officials have warned that China is launching an "AI united front" against Taiwan as the nine-in-one general election is approaching.
- China has systematically collected 23 million Taiwanese household registration and account information, used AI big data to accurately divide the audience into four major groups, and used overseas simulated accounts to deliver customized messages in an attempt to intensify social opposition and intervene in Taiwan's elections.
AI-generated summary
Why It Matters
As the 9-in-1 election is approaching, German academic institutions have revealed documents leaked from Chinese companies, saying that China is systematically collecting Taiwan household registration and account information.
National security officials have warned that China is launching an "AI united front" against Taiwan as the nine-in-one general election is approaching. (Reuters file photo)
First published 11:59
Update time 18:38
With more than 40 days left until the 9-in-1 election, national security officials warned that China is launching an "AI united front" against Taiwan. China has systematically collected 23 million Taiwanese account and household registration information, and accurately divided them into four major groups through AI big data. It uses a large number of overseas simulated accounts to deliver customized messages in an attempt to intensify social confrontation and intervene in Taiwan's elections.
National security officials pointed out that China is launching systematic information attacks and operations. During the summer vacation last year, China launched a wave of systematic attacks on 1,850 candidates in Taiwan. Recently, similar operating patterns have appeared again. By raising issues, holding press conferences and third-party collaborators, they coercively spread and deepen relevant discussions.
Officials mentioned that German academic institutions have revealed that leaked documents from the Chinese company "Golaxy" show that China is systematically collecting Taiwanese account and household registration information, covering about 23 million household registration information, using AI big data to accurately classify Taiwanese audiences into different groups such as "die-hards", "favorites", "swingers", "objectives", etc., and weaves different propaganda and attack narratives for each group to conduct accurate "audience segmentation" Segmentation)" operation.
National security officials pointed out that this operating model breaks the past single-title large-scale propaganda, but uses AI to quickly create a large number of simulated accounts and platforms to conduct long-term cross-platform "account maintenance" (such as daily sharing of soft content such as the meaning of life). At critical moments, it will switch to posting political trends and offensive remarks, and a dedicated "information support unit" will continue to conduct follow-up monitoring, evaluation and adjustment.
China’s information warfare tactics against Taiwan exhibit four major characteristics: “tailor-made narrative”
National security officials have observed that the recent information war tactics against Taiwan exhibit four major characteristics. The first is "tailor-made narrative", which uses audience segmentation to deliver customized messages; the second is "personalization and stratosphere penetration", where simulated accounts usually show the same stance and preferences as the target audience, but will make extreme remarks in specific events, thereby intensifying conflicts within the camp and social antagonism.
The third is to reduce the willingness to participate in politics, targeting supporters with different positions, and reduce their psychological support and voting intention by creating a narrative that "the candidate is not good" or "efforts are useless"; the fourth is to create false information to stir up people's livelihood and social disputes, and create false information related to people's livelihood issues, such as claiming that candidates from a specific political party will "open up a large number of controversial drugs" (opening up drugs) or "relaxing the activities of eight major industries at night" (opening up the porn industry), thereby triggering emotional anxiety and backlash among parents and the general public.
National security officials explained that after verification, many fake accounts on social platforms such as Threads that appeared to be from Taiwan were actually registered with overseas phone numbers. Although the information resilience of the Chinese people has improved recently, under the attack of foreign countries through cross-platform AI automated account maintenance, cross-talk and precise psychological warfare, new types of cognitive operations have become increasingly sophisticated and require continuous monitoring and response.
What to Watch
AI outlook — possibilities, not facts
Overseas immersive accounts will continue to post political trends and offensive remarks before the election
Very likely · Within weeks
Open Questions
- What specific preventive and countermeasures are in place?
- How many actual accounts were compromised?





