Amazon-owned streaming platform Twitch has begun feeding creator content into generative AI models by default, revealing the policy not as a standalone announcement but as a newly added opt-out setting buried under account security and privacy controls. The change, confirmed on August 13, 2026, covers streams, video-on-demand, clips, highlights, chat messages, text, and images posted to channels.
During a live Q&A on Twitch’s official channel, Chief Product Officer Mike Minton told roughly 3,000 viewers that the feature was deliberately not opt-in. “If this was opt-in, nobody would opt in. That’s honestly the answer,” Minton said. Head of Community Mary Kish added that Twitch was “reacting to this community’s voice” by providing an opt-out at all, and said she had personally opted out. Minton, however, said he would not opt out if he were a streamer.
The opt-out requires several steps: users must go to channel settings, open the security and privacy tab, scroll to “Training for Generative AI,” and turn it off. Twitch Support framed the move on August 12 as simply adding a setting, without explicitly announcing that Amazon model training had begun. When a viewer asked whether their videos had already been used, Minton said he did not actually know what Amazon had done in terms of model training.
Even after opting out, Twitch may still use channel content for captions, recommendations, moderation, and other AI features, though the company says those systems do not retain content for training models that produce new material. The decision follows similar moves by Meta and LinkedIn, as AI developers face training-data shortages and turn to platforms they own. Thousands of hours of streamers’ recorded voices and faces make Twitch an unusually rich source, and creators may not realize their content is being handed over.
