Twitch allows users to opt out of AI training: Amazon's default "free-riding" on content sparks public outrage, with officials stating "no one will agree voluntarily"
Twitch announced that users can prevent Amazon from using their content to train generative AI models, changing the default data collection setting from opt-in to opt-out. This move has sparked controversy, as the previous default setting was criticized for leveraging the reasoning of "no one would actively agree" to force data collection. Although the official promise is not to sell data, users have expressed strong dissatisfaction, believing that the platform has not set an industry benchmark
Twitch announced on Wednesday that users can now prevent Amazon from using their content to train generative AI models. The platform has changed its previously default-enabled data collection option to allow streamers and viewers to actively opt out.
This feature covers live streams, clips, images, chat logs, and other content within channels. Users can opt out with one click in the channel settings. Twitch's Chief Product Officer Mike Minton stated during a live broadcast that the company "respects" users' choices.
Default Enabled "Honest Reason"
The most controversial part was Minton's explanation of "why it was default enabled":
"If it were changed to an opt-in basis, no one would agree. That's the truth."
He also emphasized that data chosen to be retained would not be sold to other companies. Community head Mary Kish directly told viewers: "We don't expect you to be happy or excited about this. I also don't expect anyone to respond positively."
She also demonstrated how to turn off generative AI training permissions on-site and pointed out that "using content to train AI has become an industry standard."
User Anger and Platform Passivity
As soon as the news broke, the live chat was flooded with criticism. Many users bluntly stated: "No one would agree voluntarily because no one wants to feed their creativity and content to AI." Others believed that Twitch had the opportunity to set an industry benchmark for "refusing default collection" but chose to compromise.
This announcement also publicly confirmed for the first time that Amazon is indeed using Twitch content to train its own AI models. The tech giant acquired Twitch for nearly $1 billion in 2014 and has been continuously increasing its AI investments in recent years, trying to catch up with competitors like Google and Meta.
It is still unclear when Amazon started collecting this data, and it is uncertain how much content has actually been used for model training. Minton himself admitted that he does not know "specifically what was used and what was not."
Industry Reality and User Discontent
In the AI arms race, training models with user-generated content from platforms has become a routine operation for tech companies. Twitch's opening of the opt-out option is essentially a passive adjustment made in response to strong user backlash, rather than a proactive protection of creators' rights.
For streamers and viewers, this at least provides a clear choice. However, for the platform, the default-enabled, remedial approach has further deepened creators' distrust regarding "content being exploited."
