AI-Generated Deepfake Videos of Chinese Celebrities Spread Online
AI-generated deepfake videos of Chinese celebrities are being sold commercially on e-commerce platforms, with merchants offering to create convincing fake videos that evade platform detection. Multiple celebrities have reported their images being misused for deceptive videos and fake endorsements.
What happened
AI-generated deepfake videos of Chinese celebrities are being created and sold on e-commerce platforms. Merchants offer to create fake celebrity videos for 100 to 1,000 yuan or more, with some claiming the videos cannot be detected by platform moderation systems.
In February 2026, actor Wang Jingong reported that his image was stolen for use in a fake advertisement where the voice and lip movements appeared indistinguishable from genuine video. Other celebrities including Deng Chao, Sun Li, Xue Kaiqi, and Wen Zhengyuan have also reported suffering negative effects from AI-generated videos.
Seedance 2.0 announced on February 9, 2026 that it was suspending its real person reference material feature, but AI-generated celebrity videos continue to spread widely online.
Reported impact
The deepfake videos risk violating celebrity likeness rights and portraiture rights under Chinese law. When videos bind celebrities to specific products, they may trigger negative public perception and infringe on the right to reputation.
The videos could be used to create false news, rumors, or misleading content. If training data contains personal sensitive information, privacy breaches may occur. The deepfakes may also be used for mockery, insult, or damage to people's images, raising ethical concerns.
Index analysis
Technical challenges limit the ability to restrict AI-generated human video creation. AI models require precise identification and differentiation between real and non-real materials, and real-time detection and interception of unauthorized generation.
Solutions require action across multiple levels: technically, improve AI recognition and detection precision for human faces, and develop stricter input filtering mechanisms. At platform level, strengthen content review, add watermarks, and clarify user agreements prohibiting unauthorized use of others' images. At legal level, complete regulations clarifying liability boundaries for AI-generated content.
Governance controls
Some AI video generation platforms have attempted protective measures such as adding watermarks and restricting input types, but effectiveness remains limited.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.