Amazon just gave Twitch creators another reason to pay close attention to their settings. Twitch AI training will now allow Amazon to use creators’ streams and other channel content to help train generative AI models unless they manually opt out. That could include hours of creators talking, reacting, gaming, interacting with viewers, and building the personalities that keep audiences coming back.
@kourthousez Twitch is using your streams to train GenAI #twitch #streaming #streamer #internet #fyp ♬ original sound – KourtHouse
Twitch says creators who opt-out can prevent their streams, VODs, clips, stream chats, pictures, and channel text from being used in future training of Amazon models designed to generate or synthesize text, audio, images, or video. The controversy is not only that Amazon wants the data. It is that Twitch decided creators should have to say no rather than Amazon having to hear yes.
During a Twitch livestream addressing the reaction, Chief Product Officer Mike Minton directly acknowledged why the company chose that route. Minton said, “Why is it not opt-in? That’s what everybody is spamming in chat. I get it. ‘Let me opt-in versus making me opt-out,’” before adding, “Well, there’s an honest answer… If this was opt-in, nobody would opt-in. That’s honestly the answer.”
That admission gets to the center of the Twitch AI training backlash. Twitch knows affirmative participation would likely be extremely low, yet the company selected a system where inaction works in Amazon’s favor. A streamer who misses the announcement, never checks the relevant privacy setting, or assumes a platform would ask before using their work for generative AI training could remain included without making an active decision.
Twitch’s own announcement made the rollout sound more like the arrival of a new privacy control than the opening of creator content to Amazon’s models. On its official support account, Twitch announced it had “[added] a setting that lets you opt-out of having your channel content used to train generative AI content models across Amazon.”
The potential harm goes beyond a philosophical debate about artificial intelligence. Streamers build businesses around recognizable voices, humor, language, reactions, storytelling styles, community interactions, and visual identities. Those qualities are not side effects of a stream. For many creators, they are the product.
When Twitch AI training can pull from hours of that material, creators have reason to ask what happens when their work becomes an input for systems capable of generating text, audio, imagery, and video. Twitch has not said in the reporting cited here that Amazon is building models specifically designed to imitate individual streamers, so there is no basis to claim that it is. The larger concern is control. Material creators spent years producing can now contribute to technology whose eventual products and commercial uses may extend far beyond the stream where that material originated.
Compensation is another unanswered issue. No public reporting describes a creator payment program specifically tied to this generative AI training. That creates an uncomfortable value exchange. Creators provide the performances and communities that make Twitch valuable, while Amazon potentially gains another use for the resulting archive without offering a separate AI training check.
There is also a problem involving viewers. According to The Verge, Twitch says that when someone participates in another streamer’s chat, that channel owner’s preference determines whether those chat messages can be used for training. In other words, the person typing a message may not control the AI training setting that governs that conversation.
And opting out does not remove creators from every form of artificial intelligence or machine learning on Twitch. The platform says features connected to areas such as recommendations, community safety, captions, and monetization assistance may continue operating even when generative AI training is disabled. The specific setting is aimed at future training of Amazon models that generate or synthesize content.
The word “future” is especially important because there is still uncertainty about what may have happened before this rollout.
When a Twitch user asked whether videos had already been used in training, Minton said, “I don’t actually know the answer to that question because I don’t know what Amazon […] has done in terms of model training and what they’ve used and not used.”
That answer leaves creators without a clear picture of whether opting out only closes the door going forward or whether some Twitch material has already entered earlier Amazon AI development work. Minton also discussed Twitch content being used in a prototyping capacity in 2024, making the lack of a definitive answer about prior training even more significant.
Twitch Head of Community Mary Kish defended the existence of an opt-out option as a response to creator concerns. Kish called it “a reflection of reacting to this community’s voice that you are not wanting to train Gen AI models, and we want to give you that option.”
Creators who want out must go into Twitch channel settings, open Security and Privacy, locate “Training for Generative AI,” and turn the option off. Twitch AI training may be presented as something creators can decline, but the bigger story is who was given the responsibility to make that choice.
Amazon wants the material. Twitch controls the setting. Yet the creators whose voices, faces, conversations, and communities supply the data are the ones required to find the switch and say no.
