Twitch creators now have a new setting that affects how their content can be used for artificial intelligence training. The platform has added an opt-out for generative AI training connected to Amazon, Twitch’s parent company, after creator concern over whether posts, streams, and videos were being used by default.
The change is simple to use, but the debate around it is not. It has raised larger questions about user data, creator control, and the pressure on AI companies to find high-quality material for model training.
What Twitch changed
Twitch has updated its account settings so streamers can disable the use of their channel content for generative AI training. The setting applies to content that could help train Amazon’s artificial intelligence models.
For creators who want to opt out, the route is direct. From the Twitch website or mobile app, users click their account avatar, open Settings, go to Security and Privacy, and find the Generative AI Training option. From there, they can turn off the use of their content for that specific purpose.
That gives streamers a visible control point. It also makes explicit that Twitch content can be part of a broader AI training pipeline unless the creator disables the setting.
The setting does not remove every use of creator content by Twitch or Amazon. Twitch notes that “disabling this option does not prevent Twitch and Amazon from using your channel’s content for other purposes described in Twitch’s Privacy Notice.”
Those other purposes include AI-powered platform features meant to support growth, monetization, discovery, and safety. The source article lists examples such as real-time help for sponsorship campaigns, recommendations that help viewers find content, and community safety tools like AutoMod.
Why creators reacted
The opt-out has reassured some creators because it gives them a clear action they can take. But the rollout also surfaced a more uncomfortable issue: it remains unclear exactly when Twitch began using posts, streams, and videos from users to train Amazon’s systems.
That uncertainty matters because many creators see their channels as both creative work and business infrastructure. Streams, videos, community posts, and channel identity can represent years of labor. When that material is used to train generative AI models, creators may reasonably want to know when the practice started, what content was included, and who benefits from the output.
In a forum focused on the topic, more than 16,000 creators expressed opposition to having their content used by default to train Amazon’s AI systems. According to the source article, the practice only became clear after the update to the settings.
The backlash followed a livestream where Mary Kish, Twitch’s head of community, explained the new changes. She acknowledged that the change would provoke a negative reaction. Mike Minton, Twitch’s head of product, also addressed the issue and described his answer as “a candid response.”
Minton said keeping the AI training option enabled by default was necessary because otherwise “no one would participate” in the process. That statement helped crystallize the creator concern: the setting is technically optional, but the default position matters because many users may never visit that part of the settings menu.
The unanswered questions
The new control answers one narrow question: creators now know where to go if they want to disable this particular AI training use. But it does not resolve the broader questions raised by the change.
The source article identifies several open issues:
- Since when has Twitch content been used to train AI models?
- Is Amazon the only company using this data?
- Are business partners also involved?
- How is authorship of creator content respected?
Minton also suggested the issue reaches beyond Twitch itself. He said these mechanisms are not unique to Twitch and considered it likely that other companies developing AI systems are extracting content from Twitch and other services to train models.
“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.”
That point makes the Twitch AI training setting both important and limited. It gives creators a way to act inside Twitch’s own system, but it does not guarantee that publicly available content has not already traveled elsewhere or been used by other AI developers.
Terms, consent, and the default setting
Twitch’s Terms of Service, in effect since March 2024, say users grant Twitch and its sublicensees rights to use, reproduce, modify, adapt, distribute, and create derivative works from their content. The key issue is that, until now, those terms did not explicitly say the materials could be used to train generative AI models.
That gap is central to the current debate. Broad platform rights are common in digital services, but generative AI training changes how creators understand reuse. A video might be distributed to viewers, surfaced through recommendations, moderated through safety tools, or used to help build systems that generate or analyze future content.
Those uses can feel very different to the people who produced the material. The new Twitch setting recognizes that creators should have a say in at least one of those uses. At the same time, making the option enabled by default has led many creators to question whether the platform is putting the burden on users to discover and reject a practice they may not have expected.
The larger AI training data problem
The Twitch case also reflects a wider pressure point in artificial intelligence development: the shortage of high-quality training data. As demand for data rises and available sources become harder to access, companies are looking for more ways to feed AI systems.
The source article notes that OpenAI has reached agreements with various publishers, including Condé Nast, WIRED’s corporate parent, to use some content for training. It also points to other examples involving Meta, Facebook and Instagram posts and images, activity and screenshots from smart glasses users, and similar cases documented involving Google and YouTube.
In that context, user-generated content becomes a strategic resource alongside chips, memory, and electricity. For platforms, the content people create every day is valuable because it is fresh, varied, and tied to real communities. For creators, the question is whether access to free digital services should also mean contributing to AI training by default.
Twitch’s new opt-out does not settle that debate. It does, however, make the tradeoff more visible. Streamers who care about Amazon AI training now have a setting to check, and the wider creator economy has another example of how generative AI is reshaping the relationship between platforms and the people who make them valuable.