Twitch said Aug. 12 that creator channel content can be used to train generative AI content models across parent company Amazon unless streamers turn off a new account setting. The default places the consent burden on creators and gives Amazon access to a potentially valuable pool of live entertainment content, audio and video for AI development.
Twitch Support announced the control Wednesday and directed users to the platform’s account-settings guide. TechCrunch independently reported that the setting permits training by default and documented a live Twitch discussion with company executives.
Creators must switch the training setting off
Creators who do not want their channel content used for Amazon’s generative AI model training must open their Twitch account settings, choose the Security and Privacy tab, find “Training for Generative AI” and disable the option, according to Twitch and TechCrunch. The setting is located in account preferences rather than the Creator Dashboard, which could make it less visible to people focused on managing broadcasts.
During an official Twitch livestream, Chief Product Officer Mike Minton acknowledged that the company chose an opt-out design because participation would be low under an opt-in approach, TechCrunch reported. Head of Community Mary Kish said the availability of an opt-out reflects the community’s opposition to generative AI training.
The policy does not establish that every creator’s material has already been added to an Amazon model. When asked during the livestream whether Twitch content had previously been used, Minton said he did not know what Amazon had used for model training, according to TechCrunch. That uncertainty leaves an important distinction between permission to use content, actual ingestion and future model deployment.
A creator platform becomes a training-data supplier
Twitch hosts long-form, unscripted broadcasts that combine speech, gameplay, on-camera performance and audience interaction. That mix can be commercially useful for developing systems that generate or interpret video, audio and conversational material. It also makes the consent question more complicated than it is on a text-only service.
Streamers often build businesses around recognizable voices, appearances, formats and communities. An account-level opt-out asks each creator to monitor a platform policy and take action to protect that body of work.
Twitch did not frame the change as compensation or a licensing program for creators. The company presented it as a privacy control. For creators, agencies and entertainment companies, that means the immediate operational task is to review the setting, document the choice and determine whether collaborators or talent agreements require a consistent policy.
Default settings are becoming AI governance decisions
The commercial consequence extends beyond Twitch. Platforms with large archives of user-generated content increasingly hold data that can improve generative systems, while the people who produced it may not regard model training as part of the original exchange. Whether a control begins on or off can therefore determine participation at enormous scale.
For Amazon, the change connects a major creator platform with a broader AI portfolio. For Twitch, it tests whether an opt-out can preserve access to training data without further damaging trust among streamers who already view generative AI as a threat to creative work and identity.
Executives overseeing communities or marketplaces should treat AI-training defaults as product, legal and brand decisions at once. A technically available opt-out may satisfy a control requirement, but placement, disclosure and timing shape whether users experience that control as meaningful consent. Twitch’s rollout offers a live case study in how quickly a settings decision can become a creator-relations issue.