Twitch turned on AI training for streamers by default this week, and the backlash was instant. The Amazon-owned platform added a setting called Training for Generative AI, and it lets your streams, VODs, clips, and chat be used to train Amazon’s generative models. Everyone is opted in until they go flip it off. The company confirmed the change on August 12, and the creator community responded fast and loud.
I’ve written about creators losing control of their content before, so this one hit close to home. The AI music story I covered earlier this month showed what happens when machines train on work people never agreed to share. This is the same fight, just moved onto the biggest livestreaming platform on the planet.
What the AI training toggle actually covers
The new control lives in your account settings under Security and Privacy, and you can reach it directly at twitch.tv/settings/security. The toggle is labeled Training for Generative AI, and it’s on for every account by default.
Twitch’s own support page spells out what opting out does. If you turn it off, your streams, VODs, clips, stream chats, and pictures and text on your channel won’t be used in future training of a model developed by Amazon whose purpose is to generate or synthesize text, audio, images, or video. That’s a clear and unusually specific promise, and it’s the part creators should actually trust.
What opting out does not cover
Here is the part that’s easy to miss. Turning the toggle off doesn’t opt you out of all AI or machine learning uses at Twitch. AutoMod keeps running, because the company says it doesn’t retain your data and use it to produce content. Captions keep running for the same reason. Twitch can still use your channel content for AI-supported features like sponsorship campaign assistance, recommendations, and community safety. The AI training toggle is a narrow switch, and Twitch is careful to say so.
There’s also a wrinkle for anyone who just chats. If you participate in chat on someone else’s stream, their opt-out preferences govern whether that chat can be used for training. Your own setting doesn’t protect you there. You have to care about who owns the channel you’re typing in, which is a wild thing to have to think about.
Why creators are mad
The anger isn’t subtle. Within hours of the announcement, nearly 14,000 people had upvoted a comment on Twitch’s support forum demanding the setting be opt-in instead of opt-out. On a stream addressing the change, Twitch chief product officer Mike Minton talked to a live audience of nearly 3,000 aggrieved users who were spamming anti-AI sentiment in chat.
Minton gave the most honest answer a platform executive has given on this topic in a while. When asked why the setting wasn’t opt-in, he said that if it was opt-in, nobody would opt in, and that’s honestly the answer. He acknowledged that almost every service in the world defaults this way, and that Twitch knows the move isn’t a fan favorite. At least the company isn’t pretending otherwise.
Minton also said opting out won’t hurt your reach. He claimed the setting has zero impact on your ability to be discovered on Twitch or anything like that. That’s worth holding them to, because discoverability fear is the number one reason creators will leave the AI training toggle alone.
The question nobody can answer yet
The uncomfortable part came when a streamer asked whether their content had already been used for training. Minton said he doesn’t actually know the answer, because he doesn’t know what Amazon has done in terms of model training and what it has used and not used. So the toggle protects your future content, but nobody inside the company can tell you what already happened to your old streams.
This isn’t brand new behavior either. Back in 2024, when Minton was Twitch’s chief monetization officer, he said Twitch had a role to play in training Amazon’s generative AI models and that Twitch content was already being used in a prototyping capacity. The infrastructure for this has been warming up for two years. The toggle is just the first time Twitch had to admit it out loud.
The bigger picture
Twitch isn’t alone here, and Minton said as much. Google has confirmed it uses YouTube content to train its AI models, including Gemini and Veo 3, and it doesn’t offer creators an opt-out. Meta trains its models on Facebook and Instagram posts and live streams without an opt-out either. On that scale, Twitch’s toggle is genuinely more transparent than what the rest of the industry offers.
Minton went further and admitted the obvious. He said almost any publicly available content is probably being used for model training in one form or another, whether permitted or not. That’s a striking thing for a chief product officer to say on the record, and it’s the real reason this story matters beyond Twitch.
The AI training pattern is the same everywhere: platforms collect content by default, train models quietly, and treat opt-out as a favor. The AI music chart story earlier this month was the same pattern on the music side. The difference here is that Twitch finally put the choice in writing, and the community is using it to force a real conversation.
How to opt out
If you stream on Twitch, go do this now. On the web, click your avatar, open Settings, go to Security and Privacy, scroll down to Training for Generative AI, and switch it off. On mobile, it’s under Profile, then Settings, then Security and Privacy. You only need to do it in one place, because the setting applies across desktop and the app.
The setting is on by default, so you’re opted in until you act. That’s the whole story in one sentence, and it’s why this isn’t a small change. Twitch decided that your streams belong to Amazon’s AI training unless you say otherwise, and it built the choice to be the exception rather than the rule. Go flip the toggle, and then tell the people you watch to flip theirs.
