Streamers on Twitch now have a way to stop their videos, streams, and posts from being used to train Amazon’s artificial intelligence systems. The platform recently added an account setting that allows users to opt out, but the change has also raised questions about how long that data has been collected without explicit consent.
To disable the feature, users can click on their avatar in the Twitch app or website, go to Settings, and then find the Security and Privacy section. There is a toggle labeled Generative AI Training. Turning it off prevents Amazon and Twitch from using the channel’s content for AI model training.
Twitch notes that disabling this option does not prevent Twitch and Amazon from using the channel’s content for other purposes described in Twitch’s Privacy Notice. Those purposes include AI powered platform features designed to help streamers grow and monetize, such as real time assistance for sponsorship campaigns, viewer discovery through recommendations, and community safety via tools like AutoMod.
The update is meant to give creators more control. However, its launch has drawn attention to the fact that Twitch and Amazon may have been using user content to train AI models well before anyone was asked. More than 16,000 creators have voiced opposition in a forum thread dedicated to the issue. The backlash followed a livestream featuring Mary Kish, Twitch’s head of community, who acknowledged the change would provoke a negative reaction. Mike Minton, Twitch’s head of product, in what he described as “a candid response,” said the default opt in approach was necessary because otherwise “no one would participate.”
Minton added that he suspects other companies are doing the same. “I don’t know for sure,” he said, “but I think it’s quite reasonable to assume that almost any publicly available content is used to train models in one way or another, with or without permission. So I think we also need to acknowledge that there’s a lot here that’s beyond even our direct control.”
The statements have sparked questions about how far back Twitch’s AI training goes, whether Amazon is the only company using the data, and whether creators’ authorship is being respected. Twitch’s Terms of Service, in force since March 2024, grant the company and its sublicensees broad rights to use, reproduce, modify, adapt, distribute, and create derivative works from user content. However, they have not explicitly stated that such material can be used to train generative AI models until now.
This situation reflects a wider industry problem: a growing scarcity of high quality data for AI training. OpenAI has struck deals with publishers such as Condé Nast, WIRED’s parent company, to use content. But available sources are shrinking as demand rises. Meta has been using Facebook and Instagram posts, as well as employee activity and smart glasses screenshots. Similar cases have emerged with Google and YouTube.
The scramble for training data has turned user generated content into a strategic resource comparable to chips, memory, and electricity. As AI developers race to keep up with market expectations, much of that demand is likely to be met with information produced by everyday users, often in exchange for free access to digital services.
