Twitch users push back on Amazon AI training setting

Twitch is facing criticism after users discovered that content from the Amazon-owned streaming platform can be used to train Amazon’s AI models unless creators opt out. The setting is enabled by default, allowing channel content generated by creators and audiences to be used for training, though Twitch said the data will not be re-sold to other companies.

Chief product officer Mike Minton said Twitch was respecting users by giving them an opt-out, but acknowledged the reason the feature was not opt-in: “If it’s opt-in, nobody would opt-in. That’s the honest answer.” Twitch said audio could be used to refine speech-to-text models, potentially improving automatic subtitles on Twitch streams and Amazon videos.

Head of community Mary Kish told viewers how to disable the option through the Streamer Dashboard by switching off “training for Generative AI” under Security and Privacy. Twitch said doing so prevents Amazon from using streams, clips, images, chats and other channel content for generative AI training, though some users reported the toggle appeared to turn back on after being disabled.

The timing of Amazon’s data collection remains unclear. Minton said he did not know whether users’ data had already been scraped for training or what Amazon had used. Amazon bought Twitch for nearly $1bn (£740m) in 2014 and runs a range of AI services as it competes with other major technology companies.