Twitch Enables Amazon AI Training on Creator Content by Default

Default Data Collection for AI Models

Twitch has implemented a new privacy setting that permits its parent company Amazon to use creator content for training generative AI models—and the feature is turned on by default for all accounts. The "Training for Generative AI" setting grants Amazon access to livestreams, video-on-demand content, clips, chat messages, images, and text from creator channels.

The setting only appeared after Twitch faced public criticism over its AI training practices. According to 404 Media, which first reported the details, Twitch's chief product officer Mike Minton deliberately chose an opt-out structure rather than requiring creators to opt in, citing expectations of low participation rates with a voluntary system.

Limited Control Over Content Usage

Creators who want to prevent Amazon from using their work for AI training must navigate to the bottom of their settings page and manually disable the option that reads "Allow your channel content to train generative AI content models at Amazon." However, disabling this setting does not provide complete protection.

Twitch's privacy notice clarifies that the platform can still apply AI to user content for recommendations, moderation, safety tools, and other operational features regardless of the training setting. Additionally, creators who disable the feature on their own channels may still have their chat messages used for training when they participate in other channels where the setting remains enabled—the channel owner controls that permission, not individual commenters.

Unanswered Questions About Implementation

Several critical details remain unclear. Twitch has not disclosed which specific Amazon AI models will use creator data, how much content has already been collected, or whether Amazon has trained models on historical content before implementing the opt-out mechanism. The company also has not announced any compensation structure for creators whose content contributes to AI model development.

Why it matters

This approach represents a broader pattern in how major tech platforms handle creator content for AI development. By making data collection the default state, companies maximize their training datasets while placing the burden on users to understand and navigate privacy settings. For content creators who rely on Twitch for their livelihood, this raises questions about intellectual property rights and whether they should receive compensation when their creative work becomes training material for commercial AI products.

The details of Twitch's AI training policy were first reported by 404 Media.