Twitch is training Amazon's AI on your streams

Twitch just quietly enrolled every streamer on its platform into Amazon's AI training pipeline.

The Amazon-owned streaming giant announced a new privacy toggle that lets creators opt out of having their channel content used to train Amazon's generative AI models. The catch? You are already opted in by default.

The brutal truth behind the default switch

Unless you dig into your settings to turn it off, Amazon can freely use your live broadcasts, past video archives, clips, highlights, images, and chat logs to train models that generate text, audio, images, and video.

During a live stream addressing almost 3,000 furious creators, Twitch Chief Product Officer Mike Minton was asked repeatedly why the feature was opt-out instead of opt-in. His response was brutally honest: if it were opt-in, nobody would ever opt in.

Making matters worse, Twitch confirmed the opt-out only applies to future training. When users asked whether Amazon had already scraped their archives, Minton admitted he actually had no idea what Amazon's AI teams have done behind closed doors.

Chatters also get zero say. If you write messages in someone else's stream, whether your words end up inside an Amazon AI model depends entirely on that specific channel owner's preferences.

Toggling the setting off does not strip AI from Twitch entirely. The company will still process stream data for real-time captions, recommendation algorithms, and safety tools like AutoMod, claiming those features do not retain content to generate synthetic media.

If you want to protect your streams, you can head to Twitch settings, open Security and Privacy, scroll to the bottom, and toggle off Training for Generative AI.

Why it matters

For streamers, we think opting out right away is an absolute no-brainer. Livestreamers give up dozens of hours of their voice, likeness, and personality every single week, and allowing a tech giant to repurpose that data into synthetic voices and video generators without direct compensation or revenue sharing is a raw deal.

For the broader industry, this move shows just how desperate Big Tech has become for clean, multimodal training data. Web scraping is becoming a legal minefield—highlighted by YouTube creators suing Apple in April over scraped videos—so companies are turning to their own platforms to extract internal user data. Meta did it with public Instagram posts, and now Amazon is mining Twitch.

Twitch's rollout feels remarkably cynical. Packaging an unannounced data grab as a gracious new privacy feature insults creator intelligence, and Minton's admission confirms Twitch knew creators would hate it. While Amazon gets a massive influx of rich audio and video data, burning creator trust like this could push top talent to reevaluate their platform loyalty.