Twitch has acknowledged that streamed content is used to train Amazon’s generative AI models by default, with users required to find and enable an opt-out setting in their security controls. The disclosure has prompted criticism because Twitch executives could not clearly explain what data Amazon has used or whether personal information was included, raising concerns about transparency and consent.

The policy may also create copyright and contractual risks, since streams can contain game footage, music and voice performances owned by others. Streamers and publishers have traditionally benefited from an informal promotional relationship, but using that material for commercial AI training could test those arrangements; Twitch’s chief product officer said an opt-in system would receive little participation.