Twitch Defaults to Using Creator Content for AI Training, Allowing Opt-Out

Twitch, the live-streaming platform owned by Amazon, has introduced a new policy that by default uses creators' broadcast content to train generative artificial intelligence models, unless creators actively choose to opt out in their settings. The policy covers vast amounts of live-stream recordings, audio, and video data on the platform, which are considered important resources for AI training.

In response to community protests, Twitch community chief Mary Kish and Chief Product Officer Mike Minton addressed questions from about 3,000 users during an official live stream. Twitch described the update as "adding a setting that allows creators to opt out of Amazon's generative AI model training," but this explanation failed to quell community discontent. Minton also acknowledged that he was not aware of the specific data Amazon had used for model training in the past.

Twitch noted that using public user content to train AI is not unique to the company, citing technology firms such as Meta that also use public data from their platforms for model training. Twitch emphasized that offering the opt-out option is a response to community feedback. Creators wishing to exit the AI training program can now locate the generative AI training toggle under the safety and privacy options in their channel settings and turn it off manually.