Twitch Now Has a Setting to Stop Amazon From Using Your Videos to Train AI

Twitch now lets users stop the platform from using their content to train Amazon's AI systems. The setting, called "Training for Generative AI," is found in account settings under the Security and Privacy tab. It was switched on by default, meaning Twitch had already enabled the use of user content for AI training, when The Verge discovered it on August 12, 2026 The Verge.

Generative AI refers to artificial intelligence that creates new content, such as writing text, making images, producing audio, or generating video. Tools like ChatGPT and Amazon's own AI models fall into this category. According to a Twitch support page, opting out means a user's live streams, saved recordings, clips, chat messages, channel images, and channel text will not be used to train future Amazon AI models that generate this kind of content The Verge.

Twitch's description of the toggle is clear about its limits, though. Turning it off does not stop Twitch and Amazon from using your channel content for other purposes listed in the Twitch Privacy Notice. Those other purposes include AI-assisted features that help streamers grow and earn money, help viewers find channels, and keep communities safe with tools like AutoMod, Twitch's automated moderation system The Verge.

AI-assisted features that viewers and streamers already use, such as captions and safety tools, will continue to work normally for anyone who opts out. The opt-out applies only to training generative AI models, not to the wider range of machine-learning systems Twitch runs across its platform The Verge.

There is one detail worth knowing if you stream or chat on Twitch. When you participate in someone else's chat, that streamer's opt-out setting controls whether your chat messages can be used for AI training. A viewer who has opted out on their own account is not protected by their own setting while chatting on a channel whose owner has left the toggle on. The control belongs to the channel owner, not to the individual viewer, when it comes to chat data The Verge.

The default-on position follows the approach most platforms have taken with AI training settings. Users who want to exclude their content have to find the setting and turn it off themselves. Twitch did place it in the Security and Privacy tab rather than hiding it in a developer or experimental page, which makes it easier to find for users who look through their settings. But because training is enabled by default, most streamers who never visit that tab will have their content flowing into Amazon's AI training systems.

The range of content covered is wide. Live streams, saved recordings, clips, chat logs, channel images, and channel text together make up a large collection of data across text, audio, and video. Twitch's live chat is especially useful for training AI models that need to understand informal, fast-moving, and culturally current language. Audio from streams and video from clips add types of data that are costly to gather elsewhere.

This is Amazon's training system, not just Twitch's. Amazon's generative AI work includes Bedrock, Titan models, and Alexa-related AI systems. Content from Twitch feeds into that larger ecosystem. A streamer who leaves the toggle on is contributing training data to Amazon's model development, not just to features built into Twitch.

The language in Twitch's toggle description deserves a close read. The toggle covers generative AI training only. It does not restrict the use of channel content for what Twitch calls AI-supported features for streamer growth, monetization, viewer discovery, and community safety. AutoMod falls outside the opt-out. So does any AI-driven recommendation system that uses streamer content to improve how channels are shown to viewers. A streamer who opts out of generative AI training is still fully part of Twitch's broader AI systems for operational and product features. The opt-out narrows one data flow, not the wider data economy of the platform.

The way chat settings work creates a situation where individual choices matter less than the streamer's. A streamer who opts out protects their own recordings, clips, and channel content from generative AI training. But every viewer in that stream's chat has their messages governed by the streamer's setting, not their own. A viewer who has personally opted out will still have their chat messages on an opted-in channel used for training. The effective coverage of the opt-out is scattered across the platform, with protection varying channel by channel depending on each streamer's individual choice.

Twitch announced the feature through its support account on X Twitch Support and documented it on its help pages Twitch Help.

The toggle is available now in account settings.