Twitch introduces opt-out for Amazon generative AI training

Twitch has updated its account settings to allow streamers to opt out of having their content used to train Amazon’s generative AI models. The parent company of the streaming platform introduced a new toggle in the Security and Privacy settings, which was previously enabled by default. This change permits creators to prevent their material from being fed into Amazon’s models, which are designed to generate or synthesise text, audio, images, or video.

The opt-out mechanism covers a broad range of content types, including live streams, on-demand versions of those broadcasts, clips, stream chat logs, and images hosted on a user’s channel. By disabling the “Training for Generative AI” toggle, streamers can ensure their content is not utilised for future generative AI training purposes.

Despite this new control, Twitch clarified that content may still be used for other machine learning applications. These include essential platform functions such as automated captions and safety tools like AutoMod. The company stated that AutoMod does not retain user data or use it to generate material, distinguishing these operational tools from generative training pipelines.

Twitch emphasised that certain AI-driven functions are critical for community safety and cannot be disabled on an individual basis. The platform noted that allowing individual streamers to turn off these specific safety features would weaken security for the wider user base. Consequently, while generative training can be opted out of, core safety mechanisms remain mandatory.

The settings also account for cross-channel interactions. If a streamer participates in another user’s chat, the host’s privacy settings will determine whether the posted content is used for generative AI training. This ensures that the rights of the primary content creator are respected even when their audience engages with their material elsewhere on the platform.