Opt out of Twitch AI training: How to prevent your streams from training generative AI
Twitch added a new control for generative AI training on August 12, 2026. This setting governs whether channel content may be used for future training of generative AI content models developed by Amazon. At launch, TechCrunch reported that creators were included by default, so anyone who doesn't want this usage should actively check the setting.
Opting out is easy once you know where it is: Open the normal Twitch account settings, select Security and Privacy, scroll down to Generative AI Training, and toggle it off. The important point is understanding the scope. Turning it off blocks future training of generative models covered by this control, but it doesn't disable every AI-powered Twitch feature, nor does it guarantee that previously used training data will be removed.
Source: pexels.com
Key Takeaways
- The setting is real and available now: Twitch announced the opt-out on August 12, 2026.
- Use the normal account settings: Go to Settings > Security and Privacy, not the Creator Dashboard.
- Turn off "Generative AI training": This prevents covered channel content from being used for future training of Amazon's generative AI content models.
- The covered content is broad: Twitch lists streams, stream chats, VODs, clips, highlights, and text or images on the channel.
- Other AI-powered Twitch features can still be used: Recommendations, AutoMod, captions, and some monetization or sponsorship features are handled separately.
- Your chat in other channels is different: Twitch states that the other channel operator's opt-out preference determines whether that chat can be used for training.
How to opt out of Twitch generative AI training
The clearest verified path is through Twitch on the web. You can also open the settings page directly at Twitch Security and Privacy Settings after logging in.
- Log in to your Twitch account.
- Click your profile picture in the upper right corner.
- Select Settings.
- Open the Security and Privacy tab.
- Scroll down until you find Generative AI Training.
- Toggle the switch off.
- Refresh the page and confirm the switch remains off.
Do not look for this control in the Creator Dashboard. TechCrunch explicitly reported that the opt-out lives in normal account settings. If you manage more than one Twitch account, repeat the check while logged into each account rather than assuming one change applies everywhere.
Source: pexels.com
The relevant Twitch control is under Security and Privacy. The exact label to look for is Generative AI Training.
What Twitch says can be used for generative AI training
Twitch's account settings FAQ describes the training category broadly. If training is allowed, the platform says the following channel content may be used:
- Your live stream;
- the chat connected to your stream;
- Your VODs;
- Your clips;
- Your highlights;
- Text on your channel; and
- Images on your channel.
This is important because "Twitch generative AI training" isn't limited to a transcript or a few text fields. Depending on what a creator publishes, a channel could contain long-form audio, video, voice, on-camera appearances, gameplay footage, chat messages, profile pictures, panel graphics, and written descriptions.
Source: pexels.com
A Twitch channel can contain hours of audio and video, plus chat, clips, and profile material. Twitch lists several of these content types within the scope of the new training control.
What happens after you turn off the setting?
Twitch says opting out means the covered content will not be used for future training of an Amazon model whose purpose is generating or synthesizing text, audio, images, or video. This is the most important practical effect of the change.
The word "future" is also a key limitation. Twitch has not published a guarantee that turning off the setting will remove information from models that may have already been trained. When discussing Twitch's policy on August 12, Chief Product Officer Mike Minton said he didn't know what Amazon had used for model training previously, according to TechCrunch. Therefore, the safest interpretation is that the control is prospective and not a retroactive deletion function.
| Activity or content | What turning off the switch does |
|---|---|
| Your streams, VODs, clips, highlights, stream chat, and channel text/images | Prevents covered content from being used for future training of Amazon's generative AI content models. |
| Viewer recommendations and discovery | May continue to work as Twitch treats these as other AI-powered uses. |
| AutoMod and community safety systems | May continue to work after opting out of generative AI training. |
| AI-powered captions and similar Twitch features | May continue; the training switch is not a global "disable AI" toggle. |
| Your messages in another channel's chat | The other channel's opt-out preference determines whether that chat can be used for training. |
| Data that may have already been used in previous model training | Twitch has not promised that the new switch will remove them from previously trained models. |
Opting out doesn't turn off every AI function on Twitch
The phrasing next to the Twitch switch explicitly separates generative model training from other uses described in the Twitch privacy policy. Twitch gives examples, including real-time sponsorship campaign support, viewer discovery recommendations, and AutoMod for community safety. Its help materials also refer to AI-powered captions.
This distinction can be easily missed. A creator might reasonably choose "not to have my channel content used to train a generative content model" while still using a platform that relies on machine learning or AI for moderation, discovery, accessibility, or operational functions. If the goal is specifically to Twitch Privacy Policy must also be reviewed.
Source: pexels.com
The generative AI training switch is a privacy control, not a full opt-out of data processing. Twitch indicates that other platform purposes described in its privacy policy can continue.
Why your chat messages in other Twitch channels are a special case
Twitch's FAQ includes an unusual but important rule: when you chat in another user's stream, that channel's opt-out preference determines whether the chat can be used for training. Turning off the switch on your own channel therefore doesn't create an account-wide, universal protection for every message you type on Twitch.
For viewers who want to minimize exposure, the practical consequence is simple: avoid typing confidential, identifying, or sensitive information in public stream chats. A channel-based training preference is controlled by the channel operator, not by every individual participant in the chat.
Was Twitch AI training on by default?
At launch, yes, according to contemporaneous reports. TechCrunch reported on August 12 that creators were included by default, and Twitch executives discussed why the design was an opt-out rather than an opt-in on an official stream. The Verge also found the new setting to be on when it reviewed the Security and Privacy page, though its initial report noted it had asked Amazon for confirmation of the default.
As platform default settings can change, do not rely on what another user sees. Open your own Twitch settings and check the switch's state directly. The setting itself is the authoritative indicator for your account.
How to check if your Twitch AI training opt-out is still active
After turning the control off, refresh the Security and Privacy page and confirm it remains disabled. It is also wise to recheck the setting after any major changes to privacy policies or account settings. Twitch only introduced this control in August 2026, so phrasing and placement may change.
If the switch is not visible, try the direct web settings first instead of the Creator Dashboard, ensure you are logged into the correct account, and check the official Account Settings Help Page. . Avoid relying on social media screenshots when the live account page is available.
How This Compares to AI Opt-Outs on Other Platforms
Twitch is part of a broader shift where platforms are adding controls for the use of user-generated content for model development. The details vary significantly by service, region, and data type. For example, Meta's controls for Instagram involve a different policy structure and, in some regions, a formal appeals process, rather than a single Twitch-like channel toggle.
When reviewing multiple accounts at once, Zerlos' Instagram AI Opt-Out Guide explains the separate steps for Meta. Do not assume an opt-out on one platform will affect another company's systems.
FAQ
Where is the setting for Twitch training for generative AI?
Open your normal Twitch account settings, select Security and Privacy, then scroll down to Training for generative AI. TechCrunch notes the setting is not in the Creator Dashboard.
What content does Twitch include in training for generative AI?
Twitch lists your stream and its chat, VODs, clips, highlights, and text or images on your channel as content that can be used when training is permitted.
Does disabling Twitch AI training remove my old data from Amazon models?
Twitch describes the opt-out as preventing future training. It has not promised that turning off the control will remove information from already-trained models, so the setting should not be treated as a retroactive deletion tool.
Will AutoMod stop working if I opt out?
No. Twitch explicitly identifies AutoMod as an example of an AI-powered community safety feature that can continue to function even if training for generative AI is disabled.
Does opting out protect messages I post in another streamer's chat?
No. Twitch says the other streamer's channel preference will determine whether that stream's chat can be used for training.
Is Twitch itself training a generative AI model on streams?
Twitch support clarified that the new control pertains to Amazon's generative AI content model training. The setting itself states it enables training of channel content for generative AI content models at Amazon.
Was the Twitch AI training setting enabled by default?
TechCrunch reported that creators were opted-in by default when the feature launched on August 12, 2026. Since default settings can change, check the current status in your own account rather than assuming it will remain the same.
Conclusion
If you do not want your Twitch channel's content to be used for Amazon's future generative AI content model training, open Settings > Security and Privacy > Training for generative AI and turn off the toggle. Then reload the page and verify it remains disabled.
The opt-out is useful but narrower than a universal privacy switch. It covers future training of generative models for the channel content Twitch lists, while other AI-powered Twitch features and other uses can continue per the privacy policy. It also does not give you control over another channel's training preference whose chat you join, and Twitch has not described it as a way to delete data from already-trained models.