Twitch has quietly begun using streamers' content — livestreams, VODs, clips, chat and even channel images and text — to train Amazon's generative AI models, and the only way to opt out is a toggle buried deep in account security settings, according to reports on August 12.
Streaming reporter Zach Bussey first flagged the new setting, which sits under Security and Privacy labeled "Training for Generative AI" and is enabled by default. Twitch's FAQ says opting out will not block AI-powered platform features such as auto-commenting, ad selection and chat moderation. The FAQ also warns that even streamers who opt out can still be swept in: if they chat in a channel that hasn't opted out, their chat content can be used for training.
The decision drew immediate backlash, and Twitch addressed it in a livestream on Wednesday — with a remarkably candid answer. Asked why the setting is opt-out rather than opt-in, chief product officer Mike Minton replied: "If it was opt-in, nobody would opt-in... that's honestly the answer. So it's going to be on by default."
Head of community Mary Kish, who said she had personally opted out, argued Twitch "can't function without some forms of AI," while Minton said he would not opt out if he were a streamer. When asked whether users could see which channels were opted in, Minton said Twitch has no plans for such a feature and admitted "feedback has been quite clear" that the community opposes the change.
Amazon has been hungry for training data as it races competitors in generative AI — and Twitch, with its millions of hours of unique video and audio, is a rich new source.
