The streaming platform Twitch, owned by Amazon, has introduced a new toggle in account settings that allows users to prohibit the use of their content for training the corporation's generative models. The key nuance: the option is enabled by default for everyone, and the platform only notified users of the ability to opt out, rather than requiring consent.

What goes into the training dataset

According to the service's updated documentation, the data pool for AI training includes not only live streams, but also their recordings, clips, text messages in chat, as well as images and text elements posted on the channel page. Disabling the setting, as emphasized, will apply exclusively to future training cycles covering the generation of text, audio, images, and video.

Notably, Twitch did not provide a clear answer on whether user materials were used for training before this toggle appeared and how much data had already been collected. Chief Product Officer Mike Minton, responding to a direct viewer question, admitted that he does not have information about what specific data Amazon used when training models. Such opacity only heightened tensions within the community.

Community reaction and admission of an "honest answer"

The announcement sparked a storm of protest. Within the first few hours, a petition demanding that the setting be made voluntary (opt-in) gathered nearly 14,000 votes on Twitch's feedback platform. To smooth over the conflict, Minton and Head of Community Mary Kish went live, attracting around 3,000 viewers. During the stream, Minton made a notable and extremely candid statement: "Why not opt-in? That's what the whole chat is writing about right now. I get it. There's an honest answer... If it were opt-in, nobody would opt in."

Kish, in turn, described the introduction of the opt-out option as a response to the audience's stance, which categorically does not want their work used to train AI models.

My analysis

This incident is a striking example of a troubling trend in the industry: companies are increasingly adopting the "opt-out" model instead of "opt-in," shifting the burden of protecting their rights onto users. From a legal standpoint, Twitch is acting within the user agreement, but from an ethical one, such an approach undermines trust in the platform. Notably, this comes amid a tightening of judicial practice: in July, a U.S. court approved Anthropic's $1.5 billion payout in a lawsuit by authors over the use of books to train Claude. The question is only when similar lawsuits will reach streaming giants and whether the Anthropic precedent will mark the beginning of a rule change for everyone.