AI deepfake video of Taiwan President Lai impersonates pro-China stance
Taiwan's National Security Bureau reported that a YouTube channel attributed to China posted a deepfake video on September 17 impersonating President Lai Ching-te expressing pro-China views. The video, part of a broader disinformation campaign, combined AI-generated synthetic media with tailored messaging to undermine public trust in democratic institutions, according to official assessments.
What happened
A YouTube channel named "AI命運共同體" (AI Destiny Community) posted a 48-second deepfake video on September 17, 2025 impersonating President Lai Ching-te and Vice President Hsiao Bi-khim in conversation, with the AI-generated voice of Lai stating "America is unable to contain China, perhaps one day they will trade us out."
The video transmitted themes of "supporting China, doubting America, doubting the military, doubting Lai" that align with Beijing's long-standing disinformation narratives against Taiwan. The channel was established in late July 2025 and accumulated over 12,000 subscribers and over 50 videos in a short period, using simplified Chinese characters consistent with official Chinese state media language and context.
Reported impact
Taiwan's National Security Bureau stated the deepfake videos combined with disinformation form a new type of security threat that systemically erodes public trust in democratic institutions, extending impact beyond individual video viewers to undermine the foundation of democratic trust.
The techniques are part of a broader pattern of hybrid warfare targeting Taiwan involving fake accounts spreading false text and video content that feed into media reports, influencing public judgment of issues and attempting large-scale manipulation of public opinion.
Index analysis
Taiwan's National Security Bureau assessed the channel clearly operates as a collaborator for China in social media based on the use of simplified Chinese characters, vocabulary, context, and narrative output matching Chinese official media.
A Vanderbilt University National Security Institute study from August 2025 found a company closely linked to China's military, Zhoukexcel (中科天璣), extracted over 5,000 Taiwanese social media accounts and built dynamic psychological profiles including political views and personal values, integrating them with generative AI for targeted disinformation campaigns during the 2024 presidential election.
Governance controls
Taiwan's National Security Bureau stated authorities must employ multi-layered strategies including technical, legal, media literacy and civic education approaches to effectively counter hostile disinformation attacks and protect democratic freedom and social resilience.
Taiwan's updated national defense handbook alerts the public that foreign hostile forces may use false propaganda during both peacetime and crises to divide and weaken the public's determination to protect itself, including through fake accounts or domestic collaborators distributing deepfake videos, selective quoting and conspiracy theories.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.