Twitch streamers are pushing back after discovering that Amazon, the platform's owner, has been using their content to train generative AI models by default, with an opt-out control that only appeared this week and applies going forward, not retroactively. The setting, labeled "Training for Generative AI," sits in each account's Security and Privacy tab and was switched on for every existing channel before creators had any way to turn it off, according to TechCrunch's reporting. Twitch sent no email and posted no advance notice; the company's support account announced the new control the same day it appeared.
Content eligible for use includes stream video and audio, VODs, clips, channel images and chat logs — potentially thousands of hours per active streamer. Because chat logs are included, a creator who opts out can still have their messages swept up in AI training if they post in another streamer's chat where the setting remains on, a wrinkle that has fueled confusion in the days since the toggle appeared.
"Nobody would opt in"
Pressed on the design during a livestream, Twitch chief product officer Mike Minton offered an unusually candid explanation for why the feature defaults to on rather than off.
If this was opt-in, nobody would opt in. That's honestly the answer.
Mike Minton, Chief Product Officer, Twitch
Minton also acknowledged he could not say what use Amazon had already made of streamer content before the opt-out existed, telling viewers, "I don't actually know the answer to that question because I don't know what Amazon has done." Twitch's head of community, Mary Kish, framed the new control as a response to creator pressure, saying the company wanted to "give you that option" to keep content out of future model training.
The backlash has been swift. Streamers, many of whom already view generative AI with suspicion, have criticized the rollout as deceptive for enrolling everyone first and asking questions later. Some have adopted the hashtag "aioptedout" to signal to chat participants that a channel's content is no longer eligible for training, an informal workaround for a setting that offers no way to flag mixed-consent chat rooms. Kotaku and other creator-focused outlets have called on Twitch to make the setting opt-in rather than opt-out, arguing the burden of protecting one's own content should not fall on the creator.
Twitch has not said whether it will change the default, and the opt-out setting does not undo any training that may already have occurred using footage collected during what the company has described as an earlier prototyping phase. For now, creators frustrated with the policy have few options beyond opting out for future content and pressuring Twitch publicly, while the broader industry fight over whether platforms need affirmative consent before using user-generated content for AI training continues to play out across social media, stock-photo and publishing companies alike.