Overview
- Twitch, which announced the change Wednesday, added a new “Training for Generative AI” setting that users must turn off in Security and Privacy to stop future use of their channel content.
- The policy covers a channel’s streams, VODs, clips, highlights, stream chat, images and any text on the channel as data that may be used to train Amazon’s generative AI models.
- The control is enabled by default, and Chief Product Officer Mike Minton said on a company stream that it was not made opt‑in because “if it was opt in, nobody would opt in.”
- Twitch and Amazon have not clarified whether previously collected Twitch content has already been used to train production models, and Twitch executives said they do not know the scope of Amazon’s past model training.
- Creators have mounted a rapid backlash, urging mass opt‑outs and tagging streams, while Twitch notes that disabling the training toggle does not stop other AI features such as AutoMod or automated captions.