Here's the thing. When a platform rolls out a major data policy change and describes it as 'adding a setting that lets you opt out,' that is not transparency — that is a changelog written by a lawyer.
Twitch, owned by Amazon, has updated its account settings to allow Amazon to train generative AI models on streamers' content by default. The data in scope is broad: streams, VODs, clips, stream chats, and pictures and text on a channel. According to a Twitch support page, opting out means that material won't be used in 'future training' of an Amazon AI model 'whose purpose is to generate or synthesize text, audio, images, or video.'
The default-on setting drew immediate backlash from a community that has been loudly skeptical of generative AI. Twitch Chief Product Officer Minton addressed roughly 3,000 users in a live stream on the official Twitch channel alongside Head of Community Kish. His explanation for the opt-out structure was, to his credit, blunt: 'If this was opt-in, nobody would opt in. That's honestly the answer.'
That candor does not make the policy less aggressive. Streamers on Twitch often record themselves for many hours per week — voice, face, gameplay, conversation. For Amazon, that is an enormous corpus of labeled, real-world audio and video. For creators, it is their livelihood being quietly redirected into a corporate training pipeline without an affirmative click.
The confusion runs deeper than the default. When a user asked whether their videos had already been fed into Amazon's models before the opt-out existed, Minton said: '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.' The CPO of the platform does not know what the parent company has already consumed. That is not a reassuring answer.
Kish noted that Twitch is not unique here — Meta uses public content from Facebook and Instagram to train its own AI models — and framed the opt-out itself as a concession to community pressure. Other AI-powered features, including automated captions and the AutoMod safety tool, remain active regardless of the generative AI training toggle.
To opt out: channel settings (not the creator dashboard) → Security and Privacy tab → 'Training for Generative AI' → toggle off. Note that if you participate in someone else's chat, their opt-out preferences govern whether your messages are used, not yours.
Voltage take: Amazon is a private company and it is entitled to set terms of service on its own platform. But 'opt-out by default' on a data grab this large is the kind of move that invites exactly the regulatory scrutiny the tech industry keeps insisting it doesn't need. The irony is that Minton's honest answer — nobody would opt in — is also the strongest possible argument for why opt-in should be the legal standard. Creators built Twitch's value. Quietly harvesting that value for a separate commercial product, without affirmative consent, is the enshittification playbook running on schedule.
