ByteDance Denies Using Extras’ Data to Train Its AI Models

What happened

ByteDance Doubao PR head Liu Xing said on Weibo that online claims about Hengdian and ByteDance using the portraits and voices of 150,000 extras to train ByteDance models were “seriously false.” He said Volcano Engine is only providing technical support, personal portraits and voices require authorization, and data collected for the project will not be used to train ByteDance models. The Hengdian film and television AI service platform project is currently at the bidding stage.

Read next

• More changes in today’s briefingGoogle DeepMind announces Genie 4 for real-time world simulation with native audio

The story so far

ByteDance denied claims that it and Hengdian would use 150,000 extras’ likenesses and voices to train its AI models, while the Hengdian film and television AI service platform project remains at the bidding stage. Liu Xing, ByteDance Doubao’s head of public relations, called the online claims “seriously false” on Weibo. He said Volcano Engine is providing only technical support, all data assets belong to Hengdian or the parties granting authorization, and use of anyone’s likeness or voice requires that person’s authorization. Liu also said no unauthorized collection was taking place and that data collected for the project would not be used to train ByteDance models.