There is an important qualification. Mike Minton, Twitch’s chief product officer, previously told The Information that Amazon’s work with Twitch material was still in prototyping and was “not in any kind of production scale” capacity. He also said Twitch had “a role to play” in Amazon’s AI development and compared the potential value of its content with Reddit data used by Google.
That makes the situation more specific than claims that every stream has already been poured into a large commercial model. Twitch has created permission to use the material and left it on by default, while publicly available information does not establish production-scale training.
Why is the permission enabled automatically?
During a community livestream, Minton gave a notably direct explanation for making the setting opt-out rather than opt-in.
“If it was opt-in, nobody would opt-in. Um, that’s honestly the answer. So it’s going to be on by default.”
The candour did not make the policy more popular. A discussion on Twitch’s Reddit community received more than 2,000 votes, while one complaint about the default setting drew more than 700 votes. Twitch’s own feedback channels also showed strong opposition.
Mary Kish, Twitch’s head of community, said she had personally opted out. She argued that Twitch could not function without some forms of AI because the technology is already used across the platform and its communities. Kish and Minton also walked viewers through the account menus, effectively confirming that the control was not exactly placed under a spotlight.
Asked why Twitch had not emailed creators, Kish called it a “good question” and said direct messages and word of mouth could spread the information further because creators do not always read email. Apparently, users missing messages is a reason to send fewer of them.
Why can chatters and guests not make their own choice?
The setting belongs to the channel owner, but the material it covers can include other people. Twitch says chat participants are governed by the streamer’s decision. Someone who opts out on their own channel may still have their messages used if they post in the chat of a channel whose owner has left training enabled.
The same consent problem can affect guests or collaborators appearing in another person’s broadcast. During the community stream, Minton was asked what happens when an opted-out creator streams alongside someone who has not opted out. He did not provide an answer, calling it a “good question.”
Twitch also has no announced plan for a public label showing whether a channel has enabled or disabled training. Chatters and guests therefore lack both direct control and an easy way to check the host’s choice before contributing material.
Minton acknowledged that community feedback was clear and said Twitch was “not tone deaf” about the reaction. The platform nevertheless kept the default unchanged. Listening and changing course remain separate product features.
Does opting out remove content already shared?
Twitch’s public explanation does not say whether switching off the setting deletes previously collected material from datasets or removes its influence from models that may already have been trained. That omission matters because future exclusion and historical deletion are not the same thing.
Amazon Web Services takes a more explicit position for its enterprise AI products. Its policy says that when a customer opts out of an eligible service, historical content previously shared for service improvement is deleted. Twitch has not publicly offered an equivalent guarantee for creators’ streams, clips, chats or images.
The platform also has not announced licensing payments, auditing rights or a process allowing creators to inspect how their work was used. Those questions have become increasingly relevant as technology companies treat user-generated material as commercially useful training data.
Following the initial reports, users continued describing inconsistent access to the control in Twitch’s mobile apps. Some Reddit users said an app update made it appear, while others still could not find it. No subsequent formal response from Twitch or Amazon clarified the mobile rollout, historical deletion or the treatment of collaborators.
For now, creators can disable Amazon generative-model training through Twitch’s security settings, assuming the option appears on their device. The larger issue remains unresolved: Twitch has decided that silence counts as permission, because asking first would probably produce an answer it does not want.
