Twitch opts streamers into Amazon AI training by default, adds opt-out

Twitch added a setting on August 12, 2026 that lets streamers opt out of having their channel content used to train Amazon’s generative AI models. It is switched on by default, so streamers are opted in unless they act.

The default is the controversy. Left on, the setting lets Amazon, which owns Twitch, use a channel’s streams, past broadcasts, clips, chat, text and images to train future generative models. The toggle sits under Settings, then Security and Privacy, as ‘Training for Generative AI.’

Twitch executives were candid about the design. ‘If this was opt-in, nobody would opt in. That’s honestly the answer,’ Mike Minton, Twitch’s chief product officer, said of the default. Mary Kish, Twitch’s head of community, compared the approach to how Meta uses public platform content for AI training.

The opt-out has limits. It only stops future training and does not touch other AI uses on Twitch, including captions, recommendations and the AutoMod moderation tool. Viewers cannot control how their own chat messages in someone else’s channel are used; that depends on the channel owner’s setting.

The move puts Twitch among the platforms treating user-generated content as training data by default, and tests how much friction creators will tolerate before the choice architecture itself becomes the story.