Twitch Faces Backlash After Making Training Amazon's AI on User Streams A Default Setting

Amazon-owned streaming giant Twitch has sparked widespread anger among creators after quietly enabling a setting that uses channel content to train Amazon’s generative AI models by default.

While users have the ability to turn the option off, the decision to automatically opt everyone in has drawn heavy criticism across the streaming community.

Executive Admits "Nobody Would Opt In"

During an official live Q&A stream addressing the changes, Twitch executives defended the default toggle while acknowledging the feature's unpopularity.

When asked directly why the feature was made opt-out rather than opt-in, Twitch Chief Product Officer Mike Minton offered a straightforward explanation: "If it's opt-in, nobody would opt-in. That's the honest answer."

Minton added that the platform was "respecting" creators by offering an exit route:

"And as I said, almost every service in the world is on by default. I think the thing that we're doing here that is unique and different is respecting your wishes to opt out of model training. And I know this is not a fan favorite. I know it's very upsetting to the community, but this is, you know, where we are."

Mary Kish, Twitch's head of community, acknowledged the pushback during the broadcast, stating: "We don't expect you to be happy or excited about this. I don't expect anyone to react to this favourably."

What Content Is Captured?

Unless a creator manually changes their security preferences, Amazon is permitted to harvest data across entire channels. According to Twitch's official support documentation, captured material can include "your streams, VODs, clips, stream chats, and pictures and text on your channel."

This data is used toward "future training" of AI systems "whose purpose is to generate or synthesize text, audio, images, or video." Twitch gave an example where user audio might be taken to "refine models that create speech to text" which could enhance automated subtitling across both Twitch and Amazon services.

Streamers who choose to disable the setting will still have access to standard community tools like AutoMod and automatic stream summaries, as these core features rely on separate machine learning mechanisms. However, if a user participates in another creator's chat room, "their opt-out preferences govern if that chat can be used for training."

How to Disable Generative AI Training

Viewers and streamers who prefer to prevent their material from being fed into Amazon’s models must manually adjust their channel preferences.

Navigate to Account Settings:

Log into your Twitch account on mobile or web and click your profile icon to open the main menu.

Access Security and Privacy:

Select the "Settings" tab, then click on the "Security and Privacy" section.

Locate the AI Training Toggle:

Scroll down toward the bottom of the page to locate the setting labeled "Training for Generative AI".

Toggle the Setting Off:

Switch off the toggle reading: "Allow your channel content to train generative AI content models at Amazon."

Community Reaction and Glitches

The rollout has triggered severe dissatisfaction across the platform's forums and social media channels. On Twitch's User Voice feedback board, a petition demanding all AI capabilities be made strictly opt-in garnered over 13,000 votes and 4,000 comments.

"On by default is criminal.... the AI narrative push is so draining," wrote one streamer online.

Another viewer pointed out during the stream: "Nobody would opt in because nobody wants to feed AI with our creativity and content."

Adding to the frustration, several creators reported technical glitches when trying to save their preferences, claiming the toggle reactivated itself automatically: "Went in and toggled the switch off, exited, went back in, switch is still on"

When questioned during the live Q&A about whether Amazon had already scraped creator streams prior to the announcement, Minton admitted he was unsure what data the parent company "has done in terms of model training and what they've used and not used."