Twitch enables Amazon AI training on creator content by default

Twitch has implemented a new policy allowing its parent company, Amazon, to use creator content to train generative AI models. The platform has introduced an opt-out setting in the Security and Privacy section of account settings, meaning the feature is enabled by default for all users.

The policy covers a wide range of channel data, including live broadcasts, video-on-demand (VODs), clips, highlights, chat logs, and any images or text posted to a user's channel. While Twitch maintains that content used for platform features like AutoMod or caption generation is not retained for AI training, the broader generative AI training remains active unless manually disabled.

The decision has sparked significant backlash from the streaming community. Many creators argue for an opt-in model, where consent is required before data usage begins. During a recent livestream, Twitch Chief Product Officer Mike Minton stated that the company chose an opt-out system because very few creators would voluntarily participate if an explicit opt-in were required. Head of Community Mary Kish also acknowledged that the change would likely provoke negative reactions.

Critics have raised concerns regarding the potential use of intellectual property, including voices and gameplay, and have questioned whether content has already been utilized for training prior to the introduction of the new setting.