Amazon Uses Twitch Streams for AI Training by Default

The default setting means Amazon can collect live broadcasts, recorded videos, clips, chat logs, channel imagery, and text from every streamer's channel. To stop this, users must navigate to their account settings, select Security and Privacy, and disable the option labeled 'Training for Generative AI.'

The policy has triggered immediate backlash across the Twitch community, with broadcasters voicing frustration over automatic enrollment, especially given many users have created anti-AI content. During a livestream announcing the update, Chief Product Officer Mike Minton addressed the criticism directly.

When users questioned why the setting was not opt-in, Minton said, 'If it was opt-in, nobody would opt in. That's honestly the answer.' He also admitted uncertainty about whether existing content had already been used, stating, 'I don't actually know the answer to that question because I don't know what Amazon has done in terms of model training and what they've used and not used.'

Twitch documentation specifies that opting out stops future channel material from training Amazon models built to generate text, audio, images, or video. However, disabling the setting does not affect operational platform tools like AutoMod or automatic closed captions, which analyze stream data in real time but do not store content for generative model training.

The change has widened the rift between digital creators and tech platforms seeking massive volumes of human data for generative AI. While developers rely on automated defaults to gather sufficient training material, creators argue their personal content and intellectual property should not be harvested without explicit consent.