Twitch creators now have a twitch AI training opt out setting for content on their channels, giving them a way to block defined material from future training of Amazon generative-AI models. The control is meaningful, but it is not a master switch for every automated system Twitch and Amazon run.
The setting, called “Training for Generative AI,” appears in the Security and Privacy section of Twitch account settings, according to reporting by The Verge, which cited Twitch support documentation. Turning it off excludes streams, VODs, clips, stream chats, and pictures and text on a creator’s channel from future training of Amazon models built to generate or synthesize text, audio, images, or video.
Twitch’s wording matters here. It refers to “future training,” so the available policy description does not establish what happens to material used before a creator changes the setting.
What does Twitch’s AI training opt out block?
It blocks Amazon from using the listed channel material in future training for content-generating models. A creator who switches it off is opting out of that stated use for their own channel’s streams and associated material.
- Live streams
- Video on demand archives, or VODs
- Clips
- Chat on the creator’s stream
- Images and text displayed on the creator’s channel
The creator’s setting also governs chat posted in that creator’s channel. That is the awkward edge case: someone who has opted out on their own channel does not control the training status of messages they post elsewhere. According to Twitch’s guidance cited by The Verge, the owner of the channel where the chat appears makes that call.
What the setting does not turn off
Amazon and Twitch say the switch does not opt a user out of other uses described in Twitch’s Privacy Notice. It also does not disable AI-supported features that remain part of operating the service.
The Verge reported that captions and safety tools continue to work after a user opts out. Twitch’s description also names tools for creator growth and monetization, including real-time sponsorship campaign assistance; viewer discovery, such as recommendations; and community-safety systems including AutoMod.
That distinction is the whole policy in miniature. The control addresses training models designed to produce new text, audio, images, or video. It does not stop Twitch from using automated systems for functions such as moderation, captions, recommendations, or the other purposes it identifies in its privacy terms.
The reporting does not settle whether the toggle is enabled by default across Twitch. The Verge found it enabled on one account and said it had asked Amazon to clarify the default. Creators who want the exclusion should check the Security and Privacy settings rather than assume their account already reflects their preference.
This story draws on original reporting from The Verge.
