Twitch finally lets streamers say no to Amazon's AI — but only if they find the switch
The opt-out arrived quietly this week, buried in security settings and switched off by default for everyone who hasn't gone looking for it.
Written by OutOfToken AI
August 13, 2026 · 4 min read · Synthesized from reporting by Ars Technica · How this works
Twitch has spent years feeding streamers' broadcasts into Amazon's AI training pipeline without asking permission. This week, the platform finally gave users a way out — but only if they know the setting exists.
A toggle nobody asked to find
The change surfaced after streaming reporter Zach Bussey posted a screenshot of a new opt-out toggle buried in Twitch's account settings. The screenshot spread fast, mostly because almost no one knew the setting had been added. Twitch itself confirmed the update in a post on X, framing it as a new courtesy for creators.
What opting out actually covers
According to Twitch's updated support page, opting out stops Amazon from using a channel's streams, VODs, clips, chat logs, and on-channel images and text for training generative AI models — systems built to produce text, audio, images, or video. Features like automated captions and safety moderation tools, which Twitch also classifies as AI-supported, keep working regardless of the opt-out choice.
"The setting is off by default — meaning every stream, clip, and chat log is fair game for Amazon's AI training unless a creator manually finds and disables it."
Years in the making, opt-out light
This isn't a new practice suddenly announced — it's an old one finally acknowledged. Ars Technica notes a Twitch executive confirmed Amazon's use of platform content for AI training more than two years ago, well before any opt-out existed. TechCrunch points out why that data is so valuable: thousands of hours of audio and video, generated by people who often talk on camera for hours at a stretch, several days a week.
The default is the controversy
The backlash isn't really about the existence of AI training — it's about consent design. 404 Media reported that the setting was opt-out rather than opt-in on its own account, and that other users online reported the same experience. Kotaku went further, describing the toggle as effectively hidden, added without meaningful warning to the millions of creators it affects.
Twitch is betting most streamers won't dig through security settings to find a switch that quietly works against their interests. Whether regulators or streamers push back hard enough to force an opt-in model instead remains to be seen — for now, silence still equals consent.
Editorial Note
All major factual claims in the article are directly corroborated by the research sources. The timeline (announcement in August 2026, Bussey's discovery, executive confirmation 2+ years prior), the scope of data affected, the default-on mechanism, and the preserved functionality of safety features are all confirmed across multiple sources. The article's critical framing of consent design is supported by 404 Media and Kotaku's reporting on the hidden/undisclosed nature of the change.
Claim Tracker
AI-assessed
Ars Technica confirms 'Amazon was using Twitch content for AI training' more than two years before the opt-out existed. TechCrunch corroborates the default-on behavior.
Kotaku documents: 'On August 12, content creator and streaming reporter Zach Bussey posted a screenshot...The screenshot quickly went viral.'
The Verge and Ars Technica both confirm the exact scope: 'your streams, VODs, clips, stream chats, and pictures and text on your channel' won't be used for generative AI training.
The Verge states: 'Other AI-supported features like captions and safety tools will still function if you opt-out of generative AI training.'
404 Media confirms: 'Twitch makes this choice for users unless they know about the new setting and proactively turn it off.' TechCrunch calls this an opt-out rather than opt-in system.
Ask AI about this story
// discussion
sign in to join the discussion