Twitch has put a switch in front of creators—but only after confirming Amazon’s AI ambitions reach into their livestreams. The dispute is less about whether AI is coming than who gets to say yes first.

The timeline sharpened on Wednesday, when Twitch announced a privacy-setting change allowing users to prevent Amazon AI models from training on their content. A separate report summed up the practical consequence: Twitch streamers “can now opt out” of Amazon AI training.¹

That announcement also marked Twitch’s clearest confirmation that Amazon is training models on livestreams. During a company broadcast, head of community Mary Kish acknowledged the policy was unlikely to land well: “We don't expect you to be happy or excited about this.”² The live chat quickly filled with anti-AI messages, while prominent streamers criticized a system in which creators are automatically enrolled unless they find and change the setting.

Twitch’s leadership framed the control as an unusually direct privacy tool in an internet where public material is routinely swept into model-development pipelines. Chief product officer Mike Minton argued that “almost any publicly available content is being used for model training in one form or another, whether permitted or not.”² For critics, that is precisely the problem: widespread industry practice is not the same as informed permission.

The company also left a key historical question unanswered. Minton said he did not know whether Amazon had already used Twitch livestreams for training, adding, “I'm not responsible for Amazon's training efforts.”² Amazon did not respond to a request for comment.

Creators can now disable the training setting in Twitch’s privacy menu. But the backlash shows the central grievance remains intact: a safeguard offered after the fact is not the same thing as consent before the data is used.