Amazon is opening Twitch’s enormous library of livestreams, videos, voices and other creator material to generative AI training by default, forcing streamers who do not want their content used to manually switch the feature off.
BBC reported that Amazon-owned Twitch will allow creators’ content to help train generative AI models across its parent company, with creators automatically included unless they opt out.
Twitch Streams Offer Amazon Hours of AI Training Material
Livestreaming gives Amazon access to a particularly rich form of training data because Twitch creators can spend hours continuously producing spoken, visual and written material.
TechCrunch described Twitch recordings as highly valuable to Amazon because they provide thousands of hours of audio and video that could be used in developing generative AI systems. Unlike a short social media post, an individual Twitch broadcast may contain extended conversations, voices, gameplay, reactions and interactions with viewers.
The controversy is therefore not simply about whether Amazon can analyze activity on Twitch. It concerns whether creative work produced for an audience on a livestreaming platform should automatically become material for training models capable of generating new text, audio, images or video.
Twitch Says an Opt-In System Would Attract Few Creators
The strongest criticism has focused on Twitch making participation the default instead of asking streamers for affirmative permission first.
Twitch Chief Product Officer Mike Minton addressed the issue during an official livestream watched by nearly 3,000 users, many of whom were posting anti-AI comments. Asked why Twitch did not make AI training opt-in, Minton gave a direct explanation: “If this was opt-in, nobody would opt in.”
That admission makes the design choice particularly significant. Twitch is effectively relying on creators taking action to withhold their material rather than requiring Amazon to obtain their approval before using it for future generative AI training.
Twitch Head of Community Mary Kish characterized the availability of an opt-out itself as a response to creators who had made clear that they did not want their work used to train generative AI.
Creators Question Whether Amazon Already Used Their Streams
The announcement also created uncertainty about what Amazon may have done with Twitch data before the new setting appeared.
TechCrunch reported that when a user asked whether Twitch videos had already been used for Amazon AI training, Minton said he did not know what Amazon had previously used or not used for model training.
That uncertainty matters because an opt-out primarily governs what happens after a creator changes the setting. Once material has already entered a model-development pipeline, simply switching off future participation may not necessarily answer questions about earlier use.
Twitch presented the announcement as the addition of a setting allowing users to opt out of having channel content train Amazon’s generative AI models, rather than leading with the fact that creators would otherwise be included automatically.
Streamers Must Manually Disable AI Training
Creators who do not want to participate need to change their privacy settings themselves.
Twitch users can open their channel settings, select Security and Privacy, locate “Training for Generative AI,” and switch the option off.
The opt-out does not necessarily mean Twitch will stop using AI elsewhere. Twitch already employs machine-learning systems for functions such as safety and content analysis, while the dispute centers specifically on channel material being used to train generative models.
Twitch’s AI Decision Reopens the Consent Debate
The backlash reflects a wider conflict between AI companies seeking enormous quantities of human-created material and creators who want greater control over how their work contributes to systems that may eventually generate competing content.
Twitch is unusual because its creators do not merely upload finished videos. Many spend hours broadcasting their faces, voices, personalities and conversations live.
Making that material available for generative AI training transforms an ordinary platform setting into a larger question about consent. Amazon gains access unless creators object—not because each streamer actively agreed to participate.
For Twitch users, the immediate choice is simple: find the new setting and decide whether to switch it off. The harder question is whether companies developing AI should have required creators to make that decision before their content became available for training in the first place.
