Twitch has introduced a new feature that allows users to object to the use of their data for AI training. The implications are disturbing.
The title image is a collage. The background image is sourced from Magda Ehlers via Pexels.
What is this all about? On August 12, 2026, Twitch introduced a new feature via their official support account on X: Streamers get the opportunity to object to the use of their content for training generative AI.
Wait a minute – does this mean that streams were previously used without the knowledge of the content creators? These and other questions were addressed by Chief Product Officer and “Money Guy” Mike Minton, as well as Community Head Mary Kish, in a Q&A session on Twitch. However, the answers seem disturbing.
On page 2, you will learn how to protect your streams from AI.
Twitch Q&A raises more questions than it answers
What were the responses? During the livestream, there were several moments when Minton and Kish may have said a bit more than they intended – while what they did not say might even be more interesting.
For instance, when a user asks how Twitch intends to handle this regarding content on its own official channels. “I think we will choose opt-out” is the rather quiet response from Mary Kish (X).
Elsewhere, a viewer wants to know if previous VODs have already been used for AI training. After a long and significant pause, Minton admits that he does not know the answer himself. He does not know what Amazon has done so far regarding model training and what data it has used for that.
According to a report by Ars Technica, Amazon, which purchased Twitch in 2014, may have been using data from Twitch to train its AI models since at least 2024.
Why is this even a topic? Aside from the fact that the use of AI is increasingly a fundamental question for many, it is especially the manner in which Twitch introduced the new feature that is causing dissatisfaction. Instead of giving users the opportunity to voluntarily provide their content, the AI training is “opt-out.”
This means in plain language: The feature is automatically activated and will remain so until you turn it off yourself. And this is despite the likelihood that many people do not want this change. The feature relies on users not noticing the introduction, being too lazy to change their settings, or simply forgetting.
Twitch is also fully aware of this, as Mike Minton is surprisingly direct in his explanation of why they decided on an opt-out model:
So why isn’t it opt-in? That’s what everyone here is spamming in the chat. I get that. […] Here is an honest answer, and I think most of you will probably appreciate it: If it were opt-in, no one would agree. That’s honestly the answer.
While this gets bonus points for honesty, it is primarily perceived by the community this way: We already know that you don’t want this; we are doing it anyway. Minton’s explanation that other services handle it the same way doesn’t help much.
What does Twitch even want to use the streams for? According to the Twitch support page, the data is primarily intended to improve AI-supported features on Amazon. So those who agree to the use of their audio data could help refine speech-to-text models that generate automatic subtitles, for example.
However, there is also talk that the AI models could generate “content” in the form of “new videos, images or text.” Exactly how this might look remains open at this point.
At Amazon Prime, several productions with AI voices have already been released, which have been met with rather mediocre reception from users: Amazon releases several anime with AI voices, which are so bad they horrify fans. Could we perhaps have heard the odd streamer involuntarily?
Your opinion is important to us!
Do you like the article? Then let us know!