Twitch made a bold move that triggered an immediate backlash across its creator community. The live streaming platform announced that parent company Amazon will automatically use streamers’ broadcast content to train generative artificial intelligence models. Instead of asking creators to give permission first, Twitch set the default training toggle to on, forcing streamers to manually opt out if they want to protect their broadcasts.

For Amazon, live stream archives hold immense value. Streamers broadcast thousands of hours of audio and video daily, giving machine learning models massive amounts of natural speech, face data, and live reactions. Twitch creators worry that automatic opt-ins mean Amazon is harvesting their personal voices, visual likenesses, and custom gaming content without their clear knowledge. Because streamers broadcast for hours every day, an automatic opt-in setting handed over vast amounts of personal media without explicit consent.

Twitch executives attempted to address community anger during an official broadcast. Head of Community Mary Kish and Chief Product Officer Mike Minton hosted a live stream to explain the policy shift. Nearly 3,000 angry viewers filled the live chat with anti-AI messages, asking why Twitch chose an opt-out approach instead of letting creators opt in voluntarily.

Minton responded directly to the chat complaints with total honesty. He stated that if Twitch made the training feature opt-in, virtually nobody would agree to participate. He admitted that Twitch leadership knew the community opposed having their media used for generative software models, which led them to choose automatic enrolment instead.

Twitch framed the change as giving creators control by adding a toggle switch inside settings, rather than admitting they opted everyone in by default. That framing created confusion among streamers, with many wondering if Amazon had already used their past broadcast data to build existing software models. When a viewer asked if past streams were fed into algorithms, Minton admitted he did not know what data Amazon used for model training in previous runs.

Kish defended Twitch by pointing out that other major tech platforms operate the same way. Meta uses public Facebook and Instagram posts to train its software models by default, leaving opt-out settings as the only choice for users outside regions with strict privacy laws. However, for professional content creators who rely on public accounts to make a living, hiding behind private settings is simply not a realistic option.

Creators who want to stop Amazon from scraping their live streams must navigate to their Twitch channel settings menu. From there, users must open the security and privacy tab, scroll down to the section labeled training for generative AI, and manually flip the toggle off.