Twitch Opts Every Streamer Into Amazon Generative AI Training by Default

Twitch has switched on generative AI training for every channel on its platform by default, giving parent company Amazon access to streams, VODs, clips, chat logs, and channel images as training material unless each user manually turns the setting off. The Amazon-owned streaming service announced the change on August 12, 2026, framing it as the addition of a new control: a setting that “lets you opt out of having your channel content used to train generative AI content models across Amazon.”

The opt-out toggle, labeled “Training for Generative AI,” sits under the Security and Privacy tab of a user’s account settings and is on for all accounts unless switched off. The announcement drew immediate backlash from streamers, and Twitch devoted part of its Patch Notes live show on the official Twitch channel later that day to defending the design.

What the Opt-Out Covers and What It Leaves Running

Twitch’s own documentation spells out both sides of the switch. Users who turn the setting off will not have their “streams, VODs, clips, stream chats, and pictures and text” used in future training of Amazon models built “to generate or synthesize text, audio, images, or video,” according to the FAQ on its account settings page.

The carve-outs are just as specific. Opting out does not cover AutoMod, the automated moderation system Twitch says does not retain user data to produce content, or AI-generated captions, which the company says do not retain streams or clips to train content-generating models. The settings description also states that turning training off “does not opt you out of Twitch and Amazon using your channel content for other purposes” under the Twitch Privacy Notice, naming sponsorship campaign assistance, viewer recommendations, and community safety as continuing AI-supported uses.

Twitch’s FAQ explains its rationale for not offering a broader opt-out: safety and security systems like AutoMod “protect the whole community,” it says, and “turning those off for one person would weaken safety and security for everyone else on the platform, so they are not individual settings.”

One Streamer’s Switch Doesn’t Cover Another’s Chat

The setting has a structural limit Twitch disclosed in its own documentation: channel-level control means chat messages follow the channel owner’s preference, not the individual user’s. “If you chat on someone else’s stream, their opt-out preferences govern if that chat can be used for training,” the FAQ states. A viewer who has opted out on their own channel still has their chat captured for training whenever they type in a channel that has left the default in place.

The same FAQ defines what is eligible when the setting stays on: “Your stream and the stream’s chat, your VODs, your Clips, Highlights, and any text or images on your Channel may be used.”

Amazon Confirmed the Training Relationship Two Years Ago

The training pipeline itself predates the toggle. Speaking at a Creator Economy Summit in 2024, Minton confirmed to The Information that Amazon was already using Twitch content to train AI models, describing it as happening “in a prototyping, not in any kind of production scale, capacity.”

“Yeah, for sure,” he said at the time. “I mean, I think obviously within the bounds of user trust within the bounds of privacy regulations, which vary all around the world, but of course we have a role to play in that.”

What changed on August 12, 2026 is not that Amazon began training on Twitch content, but that Twitch acknowledged the practice as a standing arrangement and attached a per-channel consent mechanism to it. The toggle is live now, and the default for every channel that has not touched it is on.