ByteDance Denies Using Hengdian Extras' Likeness for LLM Training
ByteDance's Doubao PR head Liu Xing officially denied rumors that the company used the portraits and voices of 15,000 Hengdian extras to train its large language model. He clarified that the partnership with Hengdian focuses on building public digital infrastructure for scene selection and production rather than training AI models on actors' likenesses. This clarification addresses growing industry anxieties regarding data privacy and the unauthorized exploitation of performers' likenesses in generative AI. It highlights how tech platforms and entertainment hubs must carefully navigate legal and ethical boundaries when deploying digital infrastructure. Volcengine acts as the technical support provider for the 'Hengdian Pi' AI platform project, ensuring that all data assets belong exclusively to Hengdian or authorized parties. The project explicitly requires individual consent for any use of personal portraits or voices, and the data is not used to train ByteDance's large models.
## BACKGROUND
Volcengine is ByteDance's cloud computing and AI service platform that provides enterprises with one-stop cloud-native technology solutions. The 'Hengdian Pi' platform was launched through a collaboration between the Dongyang Municipal People's Government and Volcengine to integrate AI into the entire film and television production workflow.