Content creators on Twitch can now choose to keep their live streams and personal channel data out of Amazon's artificial intelligence training datasets. A new privacy toggle allows streamers to opt out of having their broadcasts, video clips, channel text, and chat history used to train future generative AI models.
The new feature, labeled "Training for Generative AI," is located under the Security and Privacy tab in Twitch's channel settings. By default, the option appears to be turned on, meaning creators must manually switch it off if they want to prevent Amazon from using their content to train models that synthesize text, audio, images, or video.
Opting out of generative AI training will not affect standard platform tools. AI-supported features designed for community safety—such as AutoMod—as well as automated captions, viewer recommendations, and sponsorship tools will remain functional across all channels. However, Twitch noted that when users post messages in someone else's stream chat, the host channel's privacy preferences govern whether that chat data is used.
The update arrives as major technology platforms face increasing scrutiny from creators over data collection practices for AI development. By offering the opt-out toggle, Twitch gives streamers greater control over how their media and community interactions are used to train Amazon's generative models.