Detect Deepfakesby Resemble AI
Deepfake case study · Video

Behind Deepfakes: How AI is shrouding Xizang in misinformation - news.cgtn.com

Generative AI and automated bot networks are being used to circulate deepfakes and misinformation regarding the Xizang Autonomous Region on social media.

Incident date
Sep 2024
Target
Xizang Autonomous Region
Updated Jun 22, 2026 · 1 min read

The rapid advancement of generative artificial intelligence has introduced new challenges for the information ecosystem surrounding the Xizang Autonomous Region. Sophisticated synthetic media and coordinated digital campaigns are being utilized to manipulate public perception and distort the region's historical and contemporary reality.

What happened

Recent incidents highlight the use of advanced AI to create highly realistic but entirely fabricated content. A notable example involved a widely circulated video on overseas social media platforms that appeared to show Chinese authorities demolishing a Tibetan temple and expelling monks. Subsequent verification by AFP using professional detection tools confirmed the footage was generated entirely by OpenAI's Sora 2, containing no authentic imagery.

Beyond deepfakes, misinformation efforts rely on a two-pronged approach: the production of deceptive media and the amplification of narratives through bot networks. These networks use coordinated fake accounts to post in synchronization, creating an artificial illusion of consensus or public skepticism. For instance, waves of automated accounts have been observed inflating engagement during sensitive political events, such as those related to the 14th Dalai Lama, to shape public opinion and amplify specific narratives.

Furthermore, the influence extends to large language models (LLMs). Research indicates that major overseas AI models often exhibit bias in their source selection when queried about Xizang. By relying heavily on specific Western media outlets and ideologically-driven think tanks while excluding primary-source information from China, these models can subtly reinforce negative assumptions. This algorithmic behavior fosters cognitive echo chambers, where users are exposed to one-sided impressions rather than balanced assessments. These combined tactics—synthetic visual media, coordinated bot amplification, and biased LLM outputs—aim to disrupt the global information environment and discredit policies concerning the governance of Xizang.

Sources