Twitch to use streamer content for Amazon AI training by default, sparking community backlash
Amazon’s parent company will now utilise Twitch recordings to train generative AI models unless creators manually disable the setting, a move that has drawn sharp criticism from the platform’s user base.
Amazon has implemented a new default policy requiring Twitch streamers to manually opt out if they do not wish for their content to be used in training generative AI models for its parent company. The decision, which places the burden of consent on creators, has triggered significant backlash from the Twitch community, who argue that the lack of explicit permission and the volume of harvested personal data—including audio, video, and voice—pose risks to their livelihoods and intellectual property.
The policy change means that stream recordings, which offer thousands of hours of valuable content for model training, will be utilised by Amazon by default. Twitch executives, including Chief Product Officer Mike Minton and Head of Community Mary Kish, addressed the community’s concerns during a livestream on the official Twitch channel. Minton defended the opt-out approach by stating that an opt-in model would result in negligible participation. “If this was opt-in, nobody would opt in. That’s honestly the answer,” Minton said, acknowledging the widespread anti-AI sentiment within the chat.
In an effort to manage the fallout, Twitch framed the announcement as the addition of an opt-out setting rather than a direct declaration of AI training initiatives. This framing caused confusion among users regarding whether their content had already been used for training purposes prior to the announcement. When asked if videos had already been fed into Amazon’s models, Minton admitted he did not know the answer because he lacked visibility into what Amazon had already used for model training.
The new opt-out toggle is located in channel settings under the security and privacy tab, specifically toggling off “training for generative AI.” This setting covers streams, including on-demand versions, clips, chat logs, and images. However, opting out does not prevent the use of content for other machine learning purposes, such as automated captions and safety tools like AutoMod. Kish noted that providing this option was a reflection of reacting to the community’s voice, but acknowledged that similar practices are common across the tech industry.
Twitch’s approach mirrors broader industry trends, such as Meta’s use of public content from Facebook and Instagram for AI training. While users in the U.K. can opt out of Meta’s training, elsewhere they must use private settings, which is not feasible for monetised creators. Kish highlighted that Twitch’s inclusion of an opt-out option distinguishes it from platforms that offer no such choice, though the default opt-in status remains a point of contention for streamers who view generative AI as a threat to their work.
