Twitch turns on Amazon AI training for all user streams by default, igniting creator backlash

The policy change

On Wednesday 12 August 2026, Twitch announced on X that it had added a setting allowing users to opt out of having their channel content used to train generative AI models across Amazon. The setting is active by default, meaning that unless users manually disable it, their live streams, video-on-demand recordings, clips, highlights, chat messages, images, and text will be available for Amazon's AI training. The toggle, labeled "Training for Generative AI," sits at the bottom of the Settings page under "Security and Privacy," on both desktop and the mobile app. Twitch has not clarified whether content has already been used for training, nor when the policy takes effect.

The rationale and its limits

Mike Minton, Twitch's head of product, and Mary Kish, community director, addressed the community in a live broadcast. Minton acknowledged the unpopularity of the measure and explained why the company chose opt-out over opt-in.

“There is an honest answer. And I think most people can appreciate it: if it were optional, nobody would participate. This is the honest answer. So it's going to be on by default.”

Minton also admitted he did not know whether Amazon had already used Twitch content for AI training.

“I don't really know the answer to that question because I don't know what Amazon has done in terms of training models, as well as what they have or haven't used.”

Twitch stated that opting out will not disadvantage channels in terms of visibility, and that disabling the option does not turn off AI-powered features already on the platform, such as the AutoMod moderation tool, automatic captions, content recommendations, and ad selection. However, the opt-out has notable limits: if a user chats on another streamer's channel, that channel owner's preferences govern whether the chat messages can be used for training. The toggle is also located in account settings rather than the Creator Dashboard that streamers use daily.

Community backlash

The announcement triggered swift criticism. A UserVoice thread asking for the setting to be opt-in rather than opt-out, opened in May 2026, surpassed 13,000 votes. On Reddit, one user called the move "absolutely disgusting." British activist Ed Newton-Rex, who campaigns for AI companies to obtain creator consent before using their data, wrote on X that activating the setting by default does not constitute user consent.

“Activating this setting by default is not consent.”

Twitch confirmed it has no plans to let users see which channels have AI training enabled or disabled. Popular streamer AnEternalEnigma suggested on Reddit that the decision was imposed by Amazon, saying multiple people at Twitch internally fought against it.

Broader context

Amazon acquired Twitch in 2014. The company is not the first tech giant to use user data for generative AI training: last year, Meta announced that all public content on Facebook and Instagram, excluding private WhatsApp messages, would be used to train its own AI models unless users opted out. YouTube, by contrast, leaves the choice to creators whether to open their videos to AI companies.