Twitch Lets Streamers Opt Out of Amazon AI Training

Twitch Adds AI Training Opt-Out After Creator Backlash

Twitch has introduced account settings that allow streamers to prevent their content from being used to train artificial intelligence models owned by parent company Amazon. The change follows significant pushback from creators who only learned their streams, videos, and posts were being used for AI training after the new toggle appeared in platform settings.

More than 16,000 creators voiced opposition in a dedicated forum after discovering the practice, which had been enabled by default without explicit prior notice.

To disable AI training use, streamers can navigate to their account avatar, select Settings, then Security and Privacy, where a "Generative AI Training" option now appears. The platform clarifies that opting out does not prevent Twitch and Amazon from using content for other purposes outlined in the privacy notice, including AI-powered features for recommendations, sponsorship assistance, and moderation tools like AutoMod.

Why It Matters

The controversy underscores a broader tension in the AI industry: leading developers face acute shortages of high-quality training data, and many are turning to user-generated content from their own platforms to fill the gap. For businesses building on streaming platforms or social networks, this raises critical questions about intellectual property, data governance, and the hidden costs of "free" services. Companies that create content on third-party platforms may unknowingly be contributing to AI systems that could eventually compete with them.

Executives Acknowledge Default-On Strategy

During a livestream discussion, Twitch executives offered unusually candid explanations for their approach. Mike Minton, the platform's head of product, stated that keeping AI training enabled by default was necessary because "no one would participate" otherwise. Mary Kish, head of community, acknowledged the change would provoke negative reactions.

Minton also noted that other AI developers likely scrape Twitch content regardless of permission. "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," he said.

Unanswered Questions on Data Use

Critical details remain unclear. Twitch has not disclosed when it began using creator content for AI training, whether Amazon's business partners also access this data, or how the company respects creator authorship in training processes.

While Twitch's Terms of Service since March 2024 have granted the platform rights to use, reproduce, and create derivative works from user content, they did not explicitly mention generative AI training until now.

Part of a Larger Industry Pattern

Twitch's approach mirrors practices at other major platforms. Meta uses Facebook and Instagram posts to train its AI models and recently disclosed using employee activity and smart glasses screenshots for the same purpose. Similar data collection has been documented at Google and YouTube.

As demand for training data outpaces supply, user-generated content has become a strategic resource alongside computing chips and electricity. The pattern suggests users of free digital services are increasingly subsidizing AI development with their creative output, often without explicit awareness or meaningful choice.

These details were first reported by WIRED.