Twitch Introduces AI Training Opt-Out After Years of Data Use

Amazon's streaming platform finally offers control over AI training

Twitch announced it will allow users to opt out of having their content used to train Amazon's generative AI models, a policy change that arrives more than two years after the company first acknowledged the practice. The streaming platform confirmed in an updated support page that streams, video-on-demand content, clips, chat logs, images, and text from user channels have been feeding Amazon's AI development efforts.

According to details first reported by Ars Technica, users must actively navigate to their security settings to opt out of the program. By default, all Twitch creators remain opted into AI training unless they take action to withdraw.

The support page explains that allowing training means "your content may be used for future Gen AI model improvements." Twitch offers speech-to-text caption refinement as an example, noting that audio from streams could improve captioning both on Twitch and across Amazon's broader ecosystem.

The timeline of Twitch's AI training program

The practice dates back at least to 2024, when Mike Minton, Twitch's chief monetization officer at the time, confirmed Amazon's use of platform content during an event hosted by The Information. "Yeah, for sure," Minton said when asked directly about AI training. He added that the company operates "within the bounds of user trust within the bounds of privacy regulations, which vary all around the world."

Despite that public acknowledgment two years ago, recent online discussions among Twitch users revealed widespread uncertainty about whether their content was being used for AI development. The lack of transparency has left many creators unaware that their work has been contributing to Amazon's machine learning initiatives since the company acquired Twitch in 2014.

Twitch has not disclosed exactly when Amazon began using platform content for AI training purposes or how much data has been collected.

Why it matters

The opt-out structure places the burden on creators to protect their content rather than requiring explicit consent upfront. For the thousands of streamers who built audiences on Twitch, this means their voices, gameplay commentary, creative work, and community interactions may have already trained AI systems without their knowledge—and potentially without compensation. As generative AI tools become more capable of producing synthetic content, creators face legitimate concerns about AI-generated competitors trained on their own material. The automatic enrollment approach also raises broader questions about how platform owners leverage user-generated content for corporate AI development.

Competing interests around AI training

Streamers may object to AI training for several reasons. Some see no financial benefit from contributing to Amazon's AI capabilities. Others worry that generative AI tools trained on their content could eventually compete with or replace human creators. Still others simply prefer not to support Amazon's broader corporate interests.

Conversely, some creators may welcome the opportunity to contribute to AI advancement, particularly if it improves platform features like automated captioning that benefit accessibility.

The new opt-out mechanism gives users more control, but the default opt-in status and historical lack of transparency may leave trust damaged. Users who want to withdraw their content from AI training can do so at www.twitch.tv/settings/security.

Ars Technica first reported these details and has asked Twitch for clarification on the full timeline of AI training practices.