Twitch has confirmed that its new generative AI training system, which feeds streamer content into models used across Amazon, is switched on by default. The Twitch AI training feature covers streams, VODs, comments, and most other creator content unless a user manually opts out.
The company revealed the change indirectly, through a Twitch Support post on X pointing to a new opt-out toggle, rather than a formal blog announcement. Twitch chief product officer Mike Minton addressed the criticism during a livestream. He explained why the setting defaults to on: “If it was opt-in, nobody would opt in.” The comment confirmed that Twitch understands how unpopular the policy is with its own creator base.
Rumors of the program had circulated for weeks before the confirmation. Streamer Zach Bussey posted about it on Bluesky in July. A Twitch User Voice thread calling for all AI features to be “optional and opt-out by default” has since grown to 228 pages and nearly 14,000 votes.
How Twitch AI training works
Twitch’s account settings FAQ describes generative AI content models as tools that “can help create or synthesize content outputs like text or images when given instructions or prompts, because they are also trained on content.” According to the FAQ, allowing training means a streamer’s audio, video, and chat logs may be used for future model improvements. That includes speech-to-text systems that could improve captions at Twitch and elsewhere across Amazon.
That description sits awkwardly next to a separate reply Twitch gave to Bussey, which said “Twitch is not training GenAI models on streamer content.” Twitch has not clarified the apparent contradiction, and did not respond to a request for comment before publication.
Twitch AI training opt-out steps
Turning off Twitch AI training takes only a few clicks, though Twitch has not made the option easy to find:
- Open Twitch settings and go to the “Security and Privacy” tab.
- Scroll to the bottom of the page to find “Training for Generative AI.”
- Toggle the setting off to opt out.
Opting out does not disable every AI feature on the platform. The FAQ notes that streams, VODs, and other data can still support features such as auto-generated captions. Twitch says that data is not used to train models that generate new content.
Streamers and studios push back
Reaction on X has run heavily toward instructions for disabling the feature, alongside blunt criticism of Twitch and Amazon. Mike Futter, co-founder and analyst at game consultancy F-Squared, said the policy could reach beyond individual streamers to the developers whose games appear on stream. “What happens if a creator streams your game and is opted in? Does Twitch think it has the right to feed a developer’s game into its AI slop factory?”
Futter, who is also director of operations and publishing at indie studio Causeway Studios, said the response from game makers is likely to escalate. “I think a lot of studios are going to be talking to their lawyers over the next couple of days,” he said. “For those of us who have avoided AI, this is a clear and present danger to our businesses and our creative work.”
Twitch has reversed unpopular policy changes before, on issues ranging from sexual content rules to political speech guidelines. Whether Twitch AI training follows that pattern depends on how much pressure builds from streamers and the studios whose games appear on their channels in the coming weeks.