News

Twitch Trains AI Using Your Streams: Here’s How to Opt Out

Philipp Briel
Philipp Briel · 5 min. read
Twitch

Starting now, Twitch allows Amazon to use your streams, VODs, clips, and even your chat history to train generative AI models by default. If you don’t want this, you’ll need to actively opt out, since the new setting is enabled by default and is well hidden in the privacy settings. We’ll show you where to find it and what the opt-out actually disables.

In a nutshell:

  • Twitch has introduced a new setting that allows channel content to be used for training Amazon’s generative AI models. It is enabled by default.
  • This affects streams, VODs, clips, stream chats, as well as images and text on the channel.
  • You can disable it under Settings > Safety and Privacy > “Generative AI Training.”
  • The opt-out applies only to model training. Other AI features, such as AutoMod or automatic captions, will continue to function independently of this setting.

What Twitch Announced

According to a TechCrunch report from August 12, 2026, Twitch has rolled out a new privacy option that allows streamers to opt out of having their content used for AI training. The catch: Unless you actively opt out, the option is enabled by default, so every channel is automatically part of the training dataset.

Twitch CPO Mike Minton explained the move during a livestream with remarkable candor: “If this were opt-in, nobody would opt in. That’s honestly the answer.” Mary Kish, Head of Community at Twitch, added—according to PC Gamer—that the opt-out option was a response to the community. Given the reactions over the past few days, it’s doubtful that this has reassured many people.

Anyone who has ever streamed on Twitch knows that the platform already collects enormous amounts of video, audio, and chat data. What’s new is that Amazon now officially intends to use this data to train its own AI models—beyond the scope of the platform itself.

Here’s how to disable AI training

It takes less than two minutes to disable, but the option isn’t particularly easy to find:

In your browser:

  1. Click on your profile picture in the top right corner
  2. Go to Settings > Security and Privacy
  3. Scroll down to the “Generative AI Training” section
  4. Turn off the toggle

In the mobile app:

  1. OpenProfile > Settings > Security and Privacy
  2. Find the AI training option
  3. Turn it off

A tip that saves many users the trouble of searching: You only need to change this setting in one place; the web and app share the same account setting. According to Twitch Support (via Twitch Support on X), opting out only applies to future training; data that has already been processed will not be retroactively removed.

What content is included in the training dataset

According to information from the Twitch FAQ, as cited by 404 Media, the default setting applies to: live streams, VODs, clips, the stream chat, as well as images and text on the channel. So, if you automatically save your streams as VODs, you’re potentially giving Amazon access to your entire archive—unless you’ve opted out beforehand.

Important: Opting out only disables the training of generative AI models. Twitch itself points out that other AI-powered features, such as AutoMod for moderation or automatic caption generation, remain unaffected and will continue to function.

Massive Criticism from the Community

The reaction from the streaming community was strong. In a Twitch-owned feedback thread on the UserVoice forum, over 14,000 votes and more than 228 pages of comments were collected, demanding that AI features be offered on an opt-in basis rather than opt-out. On Bluesky, one user summed it up: “they know we don’t want it. they do not care.”

According to TechCrunch, during an official stream, around 3,000 viewers protested the announcement live in the chat. Many in the community can’t help but draw a comparison to Meta’s similar approach on Instagram and Facebook, where user-generated content is also shared by default for AI training.

There’s also pushback from the gaming industry. According to PC Gamer, consultant Mike Futter raised a question that Twitch has not yet answered: What happens if a streamer shows a game and has given their own consent, but the studio behind it wasn’t asked? Futter expects publishers’ legal departments to take a closer look at the issue soon.

Amazon’s AI Strategy Behind It

Amazon has been investing heavily in its own AI models for years—from Alexa+ to in-house image generators—and Twitch content is an obvious source of data for this: millions of hours of video, speech, and text, live and uncensored. According to 404 Media, Amazon has not just been using Twitch content since this week, but has been doing so since 2024 as part of an internal “prototyping” phase, without this being publicly disclosed. The only new development is that the practice is now official, and streamers at least have the option to object to it in the future.

We’ve already addressed how AI training and labeling requirements should generally be handled in our article on the 2026 AI labeling requirement. For platform operators like Twitch, these rules have so far applied only to a limited extent, which is further fueling the current debate.

Conclusion

It comes as no surprise that Amazon needs training data. However, Twitch CPO Mike Minton’s candid admission that an opt-in approach simply wouldn’t have yielded enough data is unusually honest and, at the same time, quite revealing. Anyone who doesn’t want their streams and chats to end up in Amazon’s training pool should take two minutes to manually disable the setting. It’s clear that users cannot rely on the platform to voluntarily respect their privacy.