Twitch Sets AI Training on Creator Content to Opt-Out

Twitch streamers have a new setting to think about, and it's on by default. Starting August 12, 2026, a toggle called "Training for Generative AI" lets Amazon use a channel's streams, clips, VODs, images, and chat to train AI models unless the streamer manually turns it off. Twitch's own chief product officer didn't try to soften the reasoning: an opt-in version, he said, simply wouldn't get used.

What Changed

Twitch had allowed some AI training on creator content before this update, but the new setting made the practice explicit and gave streamers a way to say no for the first time. The rollout announcement, posted to Twitch's official support accounts, is what put the policy in front of the wider community and set off the backlash.

Twitch's Product Chief Explains the Default

Mike Minton, Twitch's chief product officer, addressed the backlash directly during a livestream a few days after the setting appeared. Asked why the feature wasn't opt-in instead, he didn't dress it up: "If it's opt-in, nobody would opt in. That's honestly the answer," he said, adding that the setting would stay on by default. He also said the training data would not be resold to other companies, and admitted he wasn't fully certain whether archived VODs had already been used for training before the opt-out option existed.

Mary Kish, Twitch's head of community, joined Minton on the same stream and was more blunt about how the announcement was likely to land. "We don't expect you to be happy or excited about this," she told viewers.

What the Setting Actually Covers

According to Twitch's own FAQ on the feature, the primary stated use is training speech-to-text models to improve automatic captions on Twitch and other Amazon video products. But the setting's scope, as written, goes beyond audio. It covers:

• Live streams and archived VODs
• Clips
• Channel images and thumbnails
• Chat messages sent in a streamer's channel
• Other material associated with the channel

That breadth is part of why the reaction has been sharper than a typical captioning-feature rollout. Streamers pointed out that a setting broad enough to cover chat, clips, and full VODs reads more like blanket content licensing than a narrow accessibility improvement.

How to Opt Out

• Click the profile icon in the top right corner
• Go to Settings
• Open Security and Privacy
• Scroll down and uncheck "Training for Generative AI"

The catch: this only controls a streamer's own channel. Chat messages typed in someone else's channel are still governed by that channel owner's setting, not the visitor's. To work around this, streamers started tagging their channels "aioptedout" so viewers who care about the issue know which chats are actually excluded from training.

Why This Matters for YouTube Creators Too

Twitch and YouTube compete for a lot of the same talent, and plenty of channels stream on one platform while uploading long-form and Shorts on the other. That makes Twitch's opt-out default a useful comparison point, because YouTube handles the exact same decision differently.

YouTube's own third-party AI training setting, found in Studio under Channel Advanced Settings, is off by default. A channel's content is only used to train outside AI models, by companies like Amazon, Anthropic, OpenAI, or others from YouTube's approved list, if an administrator actively turns the setting on and chooses which companies to allow. Do nothing, and nothing changes.

“Opt-out by default isn't a technical detail. It's the platform deciding, on a creator's behalf, that convenience for the AI product matters more than convenience for the person whose content trains it.”

— YouTubeSkool Creator Economy Desk

Who Is Affected

• Every Twitch streamer, since the setting applies channel-wide and was switched on automatically for existing accounts
• Viewers and moderators, since chat messages sent in a channel fall under that channel's setting, not the individual's
• Multi-platform creators who stream on Twitch and also build a YouTube channel, who now have to manage two different AI training defaults across the two platforms
• Twitch/Amazon, which faces a growing, well-organized feedback thread pushing for the setting to become opt-in

What Creators Should Do Now

• Go into Twitch Settings > Security and Privacy and check whether "Training for Generative AI" is on, since it defaults to enabled
• Turn it off per channel if you don't want streams, clips, VODs, images, or chat used for AI training, and remember alt or team channels need the same change individually
• If you run a Discord or community space tied to your stream, let regular chatters know the setting doesn't protect their messages in other people's channels, only in yours
• Cross-check your YouTube third-party training setting in Studio too. It's off by default, but it's worth confirming it hasn't been changed, especially on channels managed by a team
• Add your vote to Twitch's public feedback thread if you want the default flipped, since that's the closest thing to a formal channel Twitch has offered for pushback so far

What's Next

Twitch has not announced a timeline for revisiting the default, and Minton's comments suggest the company sees opt-out as the only version of this feature that gets meaningful data. Expect the pressure campaign around the 15,000-vote feedback thread to keep building, and expect other platforms watching the reaction, YouTube included, to note how loudly creators pushed back when consent wasn't the starting point. YouTubeSkool will follow up if Twitch changes the setting or if YouTube adjusts its own AI training terms in response to the broader industry conversation.