The Twitch streaming service has added a special switch to user account settings that allows you to prohibit the use of channel materials for training generative models from Amazon. By default, this option remains active for all users, and the platform has only notified the audience of the opportunity to opt out. The corporation's training sample includes live broadcasts, recordings, short clips, chat messages, as well as images and text from the channel's page. According to the official help of the service, disabling this setting applies exclusively to future training cycles of Amazon models, which are designed to generate text, audio, images and videos. Twitch representatives did not specify whether the training began before the appearance of this switch and exactly how much material the corporation had already collected. When asked directly by one of the viewers about this, Mike Minton, the service's product director, replied that he did not have accurate information about what data Amazon used to train its neural networks.

This announcement provoked a sharp reaction from the streamer audience. To clarify the situation, Mike Minton and Mary Kish, head of community relations, went on a special broadcast on the official Twitch channel, where about three thousand viewers watched their explanations. Minton directly answered the question of why the platform did not choose a voluntary connection model, which users actively wrote about in the chat. He noted that if explicit consent was needed, none of the streamers would simply connect to the program. In turn, Mary Kish called the very appearance of the opt-out option a reaction to the position of the community, which does not want to train commercial generative models.

In the very first hours after the announcement, the proposal to make the setting exclusively voluntary gained almost fourteen thousand votes on the official Twitch feedback platform. Earlier in July, an American court approved the payment by Anthropic of one and a half billion dollars in a class action lawsuit by authors whose books were used to train the Claude chatbot.