Twitch quietly switched on a setting that lets Amazon train generative AI on streamers’ broadcasts, VODs, clips, chat, and images. It’s on by default, buried in privacy settings, and a Twitch exec admitted the reason: “If it was opt-in, nobody would opt in.”
Twitch has started feeding streamers’ content into Amazon‘s AI, and it turned the setting on for everyone without asking first.
On August 12, the Amazon-owned platform added a toggle labeled “Training for Generative AI” under a channel’s Security and Privacy settings. It’s switched on by default. Leave it on, and Twitch can use your channel’s livestreams, VODs, clips, chat messages, images, and text to “train generative AI content models at Amazon.” There was no email, no notice at login, just a new default that streaming reporter Zach Bussey happened to spot and post about, at which point it went viral.
The way it was rolled out is a large part of why creators are angry.
The quote that says it all
Twitch’s product chief didn’t dress up the reasoning. He said the quiet part plainly.
During a livestream addressing the backlash, Chief Product Officer Mike Minton was asked why the setting isn’t opt-in.
“If it was opt-in, nobody would opt in,” he said. That’s a candid admission that the default-on design exists precisely because Twitch expects most people wouldn’t agree to this if actually asked, which is the entire objection creators are raising.
Placement matters too. The toggle sits in privacy settings rather than the Creator Dashboard, where streamers actually manage their channels, so most creators will never stumble across it in normal use.
Two catches worth knowing
Opting out helps, but it doesn’t fully close the loop, and there are wrinkles.
First, the setting isn’t retroactive. Turning it off stops future training, but Minton acknowledged that Amazon had already been using Twitch data to train AI as far back as a 2024 prototyping phase, and he couldn’t say exactly what had been used. Content already ingested can’t be pulled back out of a model.
Second, your chat isn’t fully in your control. If you type in someone else’s stream and that streamer hasn’t opted out, your messages can be used for training under their setting, not yours. So opting out on your own channel doesn’t protect what you say in everyone else’s.
To be clear about the limits: opting out doesn’t disable Twitch’s other AI features, AutoMod, auto-captions, ad tools, and Twitch says those analyze streams in real time without storing content to train generative models.
Why it matters beyond Twitch
This lands in the middle of a much larger fight over AI and consent.
Creators, especially in communities that have loudly opposed AI, are being enrolled by default into training the exact technology many of them object to, and the models don’t stay on Twitch, Amazon can use them across its business. It’s the same core tension showing up everywhere in 2026: platforms treating the content people upload as raw material for AI, and asking forgiveness through a buried toggle rather than permission through a clear choice.
How to opt out is simple: Security and Privacy settings, turn off “Training for Generative AI.” Whether that should have been the default, rather than something Twitch quietly switched on and hoped nobody would notice, is the question its own product chief accidentally answered. If the honest reason to make something opt-out is that people would say no, that tells you how they’d have answered if anyone had asked.
Article compiled and edited by Derek Gibbs (entertainment editor) and the Clownfish TV newsroom.
D/REZZED is part of Clownfish TV. For more news, views, and rants on gaming, tech, and pop culture, visit clownfishtv.com. Watch the show on YouTube at @ClownfishTV where new episodes drop daily. Subscribe to the Clownfish TV podcast on Apple Podcasts, Spotify, iHeart, and wherever else you get your podcasts. Sign up for the free newsletter at more.clownfishtv.com.
Hat Tips:
Kotaku (August 2026), verified for the default-on generative-AI training setting, the covered content (streams, VODs, clips, chats, images, text), the buried opt-out location, Zach Bussey first spotting it, and opting out not disabling AutoMod, captions, or ad selection
TechTimes (via TechCrunch) (August 2026), verified for the August 12 activation, the non-retroactive nature, Minton’s inability to confirm what Amazon already used, the 2024 prototyping phase, and the opt-out-outside-the-Creator-Dashboard placement
American Bazaar (August 2026), verified for the “if it was opt-in, nobody would opt in” quote, the automatic-enrollment backlash, and the Security-and-Privacy settings path
The Register (August 2026), verified for the models being usable across Amazon beyond Twitch, the speech-to-text example, and the chat catch where another channel owner’s setting governs your messages