Twitch has recently modified its account settings, allowing streamers to opt out of their content being used to train AI models owned by its parent company, Amazon. This change comes amid increasing concern among thousands of Twitch users who question why their streams and videos were being utilized for AI training in the first place.
To opt out, users can navigate to their account settings on Twitch. After clicking on their avatar, they must select Settings, go to Security and Privacy, and find the Generative AI Training option to disable it. However, it’s important to note that opting out does not prevent Twitch and Amazon from using content for purposes stated in Twitch’s Privacy Notice, which may include AI features aimed at enhancing streamers’ growth and safety.
This update seeks to address creators’ rights regarding their content, but the implementation raises significant questions about how Twitch and Amazon previously used that content. A forum discussion revealed that over 16,000 creators expressed strong opposition to the default setting that allowed content usage for AI training without explicit consent. This backlash intensified following remarks from Twitch executives, including Mary Kish, the head of community, who acknowledged that the changes might trigger dissatisfaction.
Twitch’s head of product, Mike Minton, noted that keeping the content usage option enabled by default was essential to ensure participation in the AI training process. He indicated that this practice is not exclusive to Twitch, implying that other tech companies might also extract user-generated content for similar purposes, raising concerns about the control creators have over their own material.
Historically, Twitch’s Terms of Service authorized the use, reproduction, and modification of user content but did not clarify that it could be used for training AI models. The situation reflects a broader challenge faced by AI developers: a scarcity of high-quality data necessary for training models. While some companies secure content usage agreements with publishers, many are turning to user-generated content as a resource.
This phenomenon is mirrored in practices by companies like Meta, which utilizes user posts and images from Facebook and Instagram for AI development, and similar instances have been observed with Google.
As the demand for training data continues to rise, the ethical implications surrounding its use grow ever more pressing. The outcome of this case may set significant precedents regarding user consent and data ownership in the realm of artificial intelligence.
For further information, you can explore Twitch’s Privacy Notice and Terms of Service.
