Twitch has announced an opt-out mechanism for users who do not want Amazon to use their channel content—streams, clips, chats, images, and text—to train Amazon's generative AI models.
The feature was announced today and follows more than two years of Twitch using such content for AI training without an explicit user choice.
Creators must manually visit their security settings to opt out; they are opted in by default.
What happened
Twitch announced today that users can now opt out of allowing Amazon to use their streams, VODs, clips, chats, and channel images/text to train Amazon's generative AI models. Users must visit www.twitch.tv/settings/security to opt out; they are automatically opted in by default.
Why it matters
Twitch has been using user content for Amazon AI training for at least a couple of years without an explicit opt-out option, as Twitch's then-chief monetization officer Mike Minton confirmed in 2024. The opt-out now gives creators control over whether their content trains models that could improve captions and other Amazon services, though the default remains opted-in.
What to watch
Users must take action themselves to opt out—there is no change to the automatic opt-in status. The support page lists specific content types subject to training (streams, VODs, clips, chats, images, and text).
Twitch announced today that it is giving users the ability to opt out of Amazon's use of their channel content for generative AI training. The change addresses a practice that has been running for at least two years without explicit user control.
According to an updated support page, Twitch users can now opt out if they do not want their streams, VODs (video-on-demand recordings), clips, stream chats, and channel images and text to be used in Amazon's generative AI model training. To opt out, users must navigate to www.twitch.tv/settings/security. The critical detail is that users are automatically opted in—they must take action to decline participation.
The scope of Amazon's use is broad. Twitch's support page explains that when content is used for training, it may help improve generative AI models across multiple applications. The page provides a concrete example: audio from Twitch streams might be used to refine models that perform speech-to-text conversion, which could improve captions both on Twitch and across Amazon's services. The training applies to models whose purpose is to generate or synthesize text, audio, images, or video.
This announcement comes more than two years after the practice was confirmed. In 2024, at an event held by The Information, Mike Minton, who was Twitch's chief monetization officer at the time, was asked directly whether Amazon uses Twitch to train AI models. Minton answered affirmatively: "Yeah, for sure." He added that the use operates "within the bounds of user trust" and "within the bounds of privacy regulations, which vary all around the world," acknowledging Twitch's responsibility in the matter. Despite this acknowledgment, no opt-out mechanism existed until today.
Twitch's announcement marks the first explicit user control over AI training practices that have been in effect for over two years. In 2024, when asked directly at an event held by The Information whether Amazon uses Twitch to train AI models, Mike Minton (Twitch's chief monetization officer at the time) confirmed it plainly, adding that the use is bounded by "user trust" and "privacy regulations." However, users had no way to decline until now.
The opt-out mechanism addresses a significant gap between stated principles and user agency. Twitch frames the use as mutually beneficial—the support page notes that audio training can improve speech-to-text models, which benefit both Twitch captions and Amazon's broader services. Yet the fact that opt-out is necessary rather than opt-in, combined with the two-year delay between practice and transparency, underscores that Twitch and Amazon treated content use as a default permission rather than a choice requiring affirmative user consent.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
Anthropic's Claude AI, working on the unsolved mathematics problem known as the Riemann hypothesis, initially…
DeepSeek released V4 Pro 0813, its latest Pro model, available through OpenRouter via API
Designers Isaque Seneda and Gabriel Abrucio created ShieldFont, a font that uses ligatures to replace common w…
Attackers compromised LiteLLM, an open source AI development tool, and extracted terabytes of credentials—incl…
The Trump administration has developed an AI framework requiring federal safety testing of the most powerful U…
Anthropic added invisible watermarks to Claude's outputs to comply with the EU AI Act's requirement that AI-ge…
The AI news that matters, in one minute each morning.
Sign up free