Amazon has begun using Twitch creators' live audio and video content by default to train its generative artificial intelligence models, as reported by Ars Technica. The platform automatically includes creator broadcasts in Amazon's training data unless users manually change their channel settings.
Twitch presented the update to users by announcing a new control setting that allows streamers to opt out of generative AI training across Amazon. Following the change, Twitch Chief Product Officer Mike Minton and Head of Community Mary Kish held a broadcast on the official Twitch channel to address nearly 3,000 live viewers, many of whom posted anti-AI messages in the chat.
During the stream, Minton directly addressed user complaints regarding why the training program was not built as an opt-in feature. Minton said that if participation required creators to actively opt in, nobody would choose to opt in.
Platform Policies
Questions arose during the broadcast about whether Amazon had already processed historical stream recordings before the opt-out control appeared. Minton said he did not know what Amazon had done regarding model training or what specific material had been used.
Kish pointed out that other major services collect public user data for artificial intelligence development. Kish cited Meta, which uses public Facebook and Instagram content to train its AI models, noting that users outside the United Kingdom cannot opt out without making their profiles private.
To stop Amazon from using stream recordings for generative AI training, creators must navigate to their channel settings rather than the creator dashboard. Within channel settings, users select the security and privacy tab, scroll down to the training for generative AI setting, and toggle the option off.
