Twitch Makes Streamer Data Available for Amazon AI
Twitch now lets creators refuse generative-AI training, but its default setting leaves eligible channel content available to Amazon.
Training is on unless creators object
Twitch has added a privacy control allowing streamers to stop their channel content from being used to train Amazon’s generative-AI models. The consequential detail is the default: creators must switch the setting off if they do not want eligible broadcasts, videos on demand, clips, channel text and associated chat messages included in training.
The control appears under “Training for Generative AI” in Twitch’s security and privacy settings. Twitch says changing it prevents future use of a channel’s content for generative-model training and triggers deletion of copies previously supplied for that purpose. For messages written in another creator’s chat, the channel owner’s preference governs whether the material is eligible. The opt-out does not stop Twitch from applying machine learning to recommendations, moderation, sponsorship tools or other platform functions covered by its broader privacy terms.
Twitch also says it has been providing content to Amazon for years, meaning the new development is the availability of a user control rather than the beginning of training itself. That history makes the deletion promise and its operational scope particularly important. The company has not publicly itemized which Amazon models used Twitch data, how much content was transferred or how individual creators’ material influenced those systems.
Why it matters
Livestream archives combine several forms of scarce training material: long-form video, synchronized speech, gameplay or real-world action, creator commentary and audience conversation. They are potentially valuable for multimodal models that must understand events over time, but they also contain faces, voices, copyrighted media and contributions from people who never negotiated directly with Amazon. Twitch’s policy therefore becomes a major test of whether platforms can repurpose creator ecosystems for AI through default consent. The opt-out gives creators a practical lever, yet it leaves the burden on individuals and is unlikely to settle wider disputes over compensation, informed consent and the rights of chat participants.