Gaming3 MIN

Twitch is training Amazon's AI on your streams and chat: here's how to stop it

Streams, VODs, clips, images, text and chat conversations can all be used to improve Amazon's generative models. The setting is enabled by default, but switching it off takes less than a minute.

Twitch logo displayed on a phone resting on a computer keyboard
Image: NurPhoto / Getty Images
01

The opt-out exists. The permission is switched on by default

Twitch has added a setting that stops channel content from being used to train Amazon's generative models. The news arrived in a rather revealing way: Twitch Support announced the opt-out, confirming in the process that permission had been enabled by default.

The list goes well beyond whatever a streamer says on camera. It covers the live broadcast, its chat, VODs, clips, highlights and any text or images published on the channel. There is some fine print too: if you post in somebody else's chat, that channel's choice applies, not yours.

Twitch insists that Twitch itself is not training these models; Amazon is. That distinction may matter on an organisation chart, but it changes very little for creators whose voice, face and community conversations can help improve systems that generate text, audio, images or video inside the company that bought Twitch in 2014.

Official Twitch artwork featuring its logo, a controller, a live button and digital elements
Image: Twitch
02

How to stop Twitch using your content for AI training

Open Twitch, select your profile picture and go to *Settings → Security and Privacy. Scroll down to Training for Generative AI* and switch it off. Twitch's official account help page says the change prevents that channel content from being used in future training.

This does not disable every AI feature on Twitch. The platform may still process data for automatic captions, moderation and safety tools. The difference is that those systems should not retain your channel to teach a model how to generate new content.

Twitch chief product officer Mike Minton offered an almost comically honest explanation for making it the default: if it were opt-in, nobody would opt in. As we saw in the copyright fight over training ChatGPT on news, the legality of training on published work is still being decided country by country. Twitch presents a more immediate problem: you can switch off training for your own channel, but if you type in somebody else's chat, you have to hope that streamer found the same toggle.

00

The conversation starts here

Sign in with a supporter account to comment. Sign in

Nobody has commented yet. Want to go first?

YOUR NEXT ROUTE

Keep following AI, creativity and content

If this story interests you, these three pieces are the best place to carry on.

OPEN THE FULL TOPIC
  1. 01Suno loses in Munich: German court rules the AI music firm violated the copyright of six songsIA · 3 MIN
  2. 02What is AI slop—and are we digging our own grave?IA · 5 MIN
  3. 03AI won't kill music. It may bury us in songs nobody cares aboutIA · 4 MIN

KEEP READING

You may also like

FRONT PAGE
Steam logo over a mosaic of games available on Valve's store01OPEN
Gaming4 MIN

Steam's AI label is already hurting games

A new study of 508,192 reviews finds a 17.9-point gap between games that disclose generative AI and those using procedural generation. For many players, the label now suggests somebody tried to save money in the wrong place.