Twitch, the largest streaming platform, has quietly activated a new option in account settings for all users. Now, every streamer formally has the ability to prohibit Amazon from using channel content to train generative models. However, the point is not the emergence of an "opt-out," but that this option is disabled by default—meaning consent to use your content is already implied, not requested.
What exactly ends up in the dataset?
Amazon's training dataset includes not only broadcast archives, but also live streams, clips, chat texts, as well as images and descriptions on the channel page. According to the help documentation, the restriction applies only to future training cycles. This means that all content collected before the setting is activated may have already been used irreversibly.
The management's reaction is telling. Product director Mike Minton, when directly asked by a viewer what data Amazon has already "fed" to its algorithms, honestly admitted: he does not know. The lack of transparency in data usage matters is an alarming signal that once again demonstrates the gap between claims of "safety" and the real practices of corporations.
Stormy community reaction
The announcement sparked a wave of criticism. Minton and community head Mary Kish went live, gathering around 3,000 viewers to answer questions. Kish called the emergence of the opt-out option "a response to the community's stance." However, the most striking moment was Minton's admission: when asked why an "opt-in" mechanism (voluntary enrollment) could not be implemented, he replied bluntly: "If it worked on an opt-in basis, nobody would opt in."
This statement is essentially an acknowledgment that the platform is aware of users' reluctance to participate in AI training, yet still uses their content by default. A petition demanding that the setting be made voluntary gathered nearly 14,000 votes within the first hours on Twitch's feedback platform.
My take on the situation
This is a classic "consent by default" model, which in jurisprudence is considered one of the most controversial practices. In the context of the recent precedent where a court ordered Anthropic to pay $1.5 billion to authors for using books without permission, Amazon's actions look like a calculated risk. Regulators are increasingly paying attention to such cases, and I would not be surprised if this episode becomes the start of a new class-action lawsuit—this time from streamers.