The Amazon-owned streaming site Twitch is now using channels to train generative AI systems, and the company’s top product boss knows it’s a controversial piece of news that most people won’t like.
By default, Twitch channels are now used to train generative AI models across Amazon, and users need to opt-out. Opting-out isn’t enough, though, as any chats on a channel that hasn’t opted out will still be used for training.
During a livestream, Twitch chief product officer Mike Minton said that if the model training was on an opt-in basis, no one would do it.
“We know many of you will opt out. Why is it not opt-in? ‘Let me opt-in versus making me opt-out.’ Well, there is an honest answer, and I think most of you can appreciate this. If it was opt-in, nobody would opt-in. That’s honestly the answer. So it’s going to be on by default,” he said. “Almost every content service in the world is on by default. I think the thing that we are doing here that is unique and different is respecting your wishes to opt-out of model training. I know this is not a fan-favorite. I know it’s very upsetting to the community. But this is where we are.”
On the Twitch website, the company said the new model-training feature uses channel content to train generative AI content models, including streams, VODs, clips, stream chats, and pictures and text broadcast to the channel. Opting-out stops Twitch from doing this on your channel, but as mentioned above, chatting on someone else’s stream will still leverage the model-training if that person has not opted out.
Additionally, anyone who opts-out to this new feature will still be part of model-training, as Twitch uses AI and machine learning for its AutoMod community-safety service and for AI-supported features like automatically generated captions. It is not possible to opt-out of these offerings, though Twitch said it does not retain user data for other types of model training.
You can’t opt-out of all of Twitch’s AI uses because some AI-supported features are not individual settings, but platform offerings, Twitch said.
“Some of it runs the service itself, the same way our servers and our code do. Some of it keeps the service secure. And some of it, like AutoMod, protects the whole community. Turning those off for one person would weaken safety and security for everyone else on the platform, so they are not individual settings,” Twitch explained.
Twitch’s own head of community, GameSpot alum Mary Kish, said she is opting-out of the new AI-training features in part because some of her community members would not join her streams if she opted-in.
Twitch’s announcement drew a lot of responses. Popular streamer PaladinAmber joked that she is going to “start doing makeup that distorts my face and speak in really cryptic terms” and “f**k with the stream” so much that whatever the AI learns will be so bad it needs to be thrown out.
People are raising other issues, like what happens if a creator streams your game and has opted-in. Or what happens for union-backed voice actors who lobbied for protections against AI usage.
The push for further AI training on Twitch is not a big surprise, as its parent company, Amazon, is heavily invested in various pillars of AI. In May this year, the company said it was investing more than $200 billion into its AI buildout in 2026 alone. Amazon is also partnered with AI juggernaut Anthropic.