Twitch adds opt-out setting for Amazon generative AI training on user content
The Facts
- Twitch added a "Training for Generative AI" setting that is enabled by default on user accounts.
- The setting allows Amazon, which owns Twitch, to train generative AI models on channel content.
- Opting out covers streams, VODs, clips, stream chats, and pictures and text on a channel.
- The opt-out toggle is located in Twitch settings under the Security and Privacy tab.
- Turning the setting off does not stop other AI-supported Twitch features described in the platform's privacy notice.
- Twitch chief product officer Mike Minton said the feature is opt-out because "if it was opt-in, nobody would opt in."
- Twitch head of community Mary Kish said she did not expect users to react favorably to the change.
- A request on Twitch's feedback forum to make the setting opt-in drew roughly 14,000 upvotes within hours.
- Streaming industry reporter Zach Bussey publicized the new toggle in a screenshot posted on August 12.
- Reports indicate Amazon had been using Twitch content for AI training for at least two years before the opt-out existed.
Context
How do Twitch users opt out?
Users go to their Twitch account settings, open the Security and Privacy tab, and uncheck the "Training for Generative AI" toggle IGN,Verge. The opt-out page is at twitch.tv/settings/security Ars Technica. Opting out applies to future training of Amazon models built to generate or synthesize text, audio, images or video Ars Technica,engadget.
What does opting out not cover?
Twitch says turning the setting off does not stop Twitch and Amazon from using channel content for other purposes described in the Twitch Privacy Notice, including AI-supported features such as real-time sponsorship campaign assistance Sportskeeda,Destructoid. Features like captions and safety tools continue to work for users who opt out Verge,XDA-Developers.
Why is this being raised as a privacy concern for streamers specifically?
Twitch hosts large volumes of long-form video and audio, which reporting describes as valuable AI training data TechCrunch,Lowyat.NET. Creators often livestream themselves and their own voices for many hours each week, and critics say the default setting means content could be used without creators realizing it NewsBytes,TechCrunch,Lowyat.NET.
Where Left and Right agree, and where they split
- Where Left and Right agree
- Minton's admission that nobody would opt in is treated by both as proof that the default-on setting exists precisely because creators would not agree to it.
- Where Left and Right split
- Whether the story is about permission never being asked before Amazon took creators' streams, or about value being taken from creators without anyone naming a price.
How left and right read it
Consent that has to be hunted down in a Security and Privacy tab is not consent at all. Amazon switched the training setting on by default, so every stream, VOD, clip, and chat message feeds its models unless a creator finds the toggle — and Mike Minton's admission that nobody would opt in concedes the whole point. Roughly 14,000 upvotes in hours say otherwise. Permission should be asked, not assumed.
The value in a livestream belongs to the person who made it, yet Amazon, which owns Twitch, has channel content — streams, VODs, clips, chats — training its models by default, and Mike Minton's remark that nobody would opt in admits creators would never trade it away for nothing. That is a price signal. So when opting out still leaves other AI features running, what is Amazon paying for what it takes?
The receipts — all 49 sources
Independent coverage (49)
Facts first. Then every angle.
The day’s biggest stories in one short brief — the facts everyone agrees on, then the competing values behind the headlines. Free in your inbox.