Tech

Twitch Introduces AI Training Opt-Out Amid Creator Backlash

Amazon-owned streaming service updates settings following revelations about default data usage, as industry leaders acknowledge the strategic scramble for high-quality training data.

Editorial persona
Owen Mercer
Markets and Finance Editor
Published
Draft
Source: WIRED · View original source
Amazon Can Use Your Twitch Content to Train Its AI—Unless You Opt Out
Streamers can now disable use of their content for generative models, but platform retains rights for other AI features

Twitch has updated its account settings to allow streamers to opt out of having their content used to train Amazon’s artificial intelligence models. The new configuration, located within the Security and Privacy section under Generative AI Training, addresses significant backlash from thousands of creators who questioned the default use of their streams and posts for machine learning purposes.

The change follows revelations that Twitch’s Terms of Service, effective since March 2024, already permitted broad usage of user content. While disabling the new option prevents use for generative AI training, Twitch and Amazon retain the right to use content for other purposes, such as platform features and community safety tools like AutoMod.

Over 16,000 creators expressed opposition in a dedicated forum after the update was revealed. The backlash arose during a livestream in which Mary Kish, Twitch’s head of community, acknowledged that the changes would provoke a negative reaction. She explained that the update aims to recognise creators’ right to decide how their produced content is used.

Mike Minton, Twitch’s head of product, stated that keeping the option enabled by default was necessary because “no one would participate” otherwise. Minton suggested that it is reasonable to assume almost any publicly available content is used to train models, with or without permission, and that this is beyond Twitch’s direct control.

The move highlights the industry-wide shortage of high-quality training data. Meta has been using posts and images from Facebook and Instagram to train AI models, while similar practices have been documented involving Google and YouTube. OpenAI has reached agreements with publishers, including Condé Nast, to use content for training.

High-quality training data is considered a strategic resource alongside chips, memory, and electricity. The story originally appeared in WIRED en Español and was translated from Spanish.

Continue reading

More from Tech

Read next: EuroBirdPortal maps bird movements across Europe
Read next: WIRED names Bose earbuds its top pick for noise cancellation in 2026 guide
Read next: Three University of Florida students charged over alleged cyber-harassment of Maddie Kowalski