Twitch quietly trains Amazon AI on your streams — now you can opt out

Twitch has finally added an opt-out setting for streamers who don’t want their content fed into Amazon’s generative AI models. The company updated a support page today confirming that streams, VODs, clips, chat messages, and channel images are all fair game for training models that generate text, audio, images, or video — unless users manually flip the switch at twitch.tv/settings/security.

The kicker? Everyone is opted in by default, and has been for at least two years. Back in 2024, Twitch’s then-chief monetization officer Mike Minton admitted at a The Information event that Amazon uses Twitch content for AI training. “Yeah, for sure,” he said, adding that it happens “within the bounds of user trust within the bounds of privacy regulations.” The support page spells out what that actually means: your audio could help refine speech-to-text models that improve captions on Twitch, but also across all of Amazon.

This lands in a different place than similar policies at other platforms. When Reddit and Stack Overflow started selling user content for AI training, they at least framed it as a revenue strategy. Twitch is treating it as a default setting buried in a security menu, which feels more like compliance-driven disclosure than genuine transparency. The timing suggests legal pressure — EU privacy regulators have been sharpening their teeth on exactly this kind of default opt-in model.

Copyright questions loom large here. Streamers routinely broadcast copyrighted music, gameplay footage, and third-party content. If Amazon trains on all of it by default, the liability chain gets murky fast. Don’t be surprised if this opt-out toggle becomes exhibit A in a future class action or regulatory inquiry. Two years of silent training doesn’t disappear because a checkbox now exists.