Taiwan cybersecurity institute flags AI deepfake tactics
台灣國家資通安全研究院警告人工智慧深度偽造技術
| TAIPEI (Taiwan News) — The National Institute of Cybersecurity has identified four major types of AI-generated videos and urged the public to stay vigilant amid a rise in scams, CNA reported Sunday.
NICS researcher Tai Yu-chen (戴毓辰) said AI-generated videos are spreading online, per CNA. He warned that deepfakes tend to emerge wherever profits can be made, as AI tools now enable face-swapping and fully synthetic video production. Based on the institute’s observations, Tai said deepfake content broadly falls into four categories. These include celebrity face-swapped pornography, celebrity impersonation scams, AI avatar-driven content farms, and reuploaded AI-generated videos. Celebrity impersonation scams often misuse the likeness of financial experts or doctors, with AI-generated speech used to promote investments or medical products, Tai explained. AI avatar content farms, meanwhile, create personalities that post automated videos at high frequency. To assess authenticity, Tai outlined three checks. These include looking for platform-mandated AI disclosure labels, conducting reverse image searches, and assessing whether the scenario is logically plausible. Pang Tsai-wei (龎在偉), another researcher, said deepfakes frequently contain inconsistencies in people or settings, such as repurposed news footage used for advertisements. He added that AI still struggles with fine details such as straight lines. Tai said visible flaws will become harder to spot as AI improves. He urged users to check for AI labels and search for original sources to avoid being scammed. |
|
| Michael Nakhiengchanh Taiwan News, Staff Writer | |
| 2026-01-26 |
