Twitch has introduced a setting allowing streamers and viewers to prevent Amazon from using their streams, clips, chats and images to train generative artificial intelligence models. The livestreaming platform confirmed the change on Wednesday, revealing for the first time that Amazon, which acquired Twitch in 2014 for nearly $1bn (£740m), had been using channel content for AI development.
According to Twitch's support account, the new setting lets users opt out of having their content used for generative AI training, confirming the feature is now live on the platform's help system. A technology outlet's reporting notes the control sits within Security and Privacy settings under a section labelled 'Training for Generative AI'.
What exactly can users opt out of?
The setting covers a broad range of channel material rather than just video. Per a technology publication's breakdown, the opt-out applies to streams, VODs, clips, stream chats, and the pictures and text displayed on a channel, while AI-supported tools like captions and safety moderation remain active regardless of the setting. Twitch has said the option is intended for models that generate or synthesise text, audio, images or video, according to a gaming news outlet.
Notably, opting out does not switch off every AI or machine-learning use on the platform. A business publication's report states that Twitch's own FAQ clarifies the setting only halts future training of Amazon's generative models, leaving existing AI-assisted features such as automatic captions and moderation tools untouched.
Why was data collection switched on by default?
Twitch's chief product officer, Mike Minton, addressed this directly during a livestream, explaining that making the setting opt-in from the start would have led to negligible participation.
If it's opt-in, nobody would opt-in. That's the honest answer.
Minton said that any user who remains opted in could have their channel content used for training, though he stressed the collected data would not be resold to other companies. He was unable to say whether Amazon had already scraped user data before the opt-out was introduced, telling viewers he did not know precisely
what they've used and not used.
How did Twitch explain the decision to users?
Mary Kish, Twitch's head of community, walked viewers through the process of disabling generative AI training via channel settings during the same broadcast. She acknowledged that data collection for AI purposes has become widespread across the industry.
We don't expect you to be happy or excited about this. I don't expect anyone to react to this favourably.
The livestream's chat filled with hundreds of critical comments as it aired. One viewer wrote:
Nobody would opt in because nobody wants to feed AI with our creativity and content.Another argued that Twitch had a chance to set a precedent for the wider industry by resisting the practice altogether.
How does this fit into the wider AI debate across platforms?
Twitch's move echoes tensions playing out elsewhere in the technology sector over AI training and consent. Meta recently faced criticism after its Muse Image tool generated pictures from public Instagram photos without users' consent, prompting the company to add its own opt-out setting. Separately, platforms including Snapchat have begun restricting the spread of wholly AI-generated video content, joining a broader industry effort to manage AI-generated material. Twitch has previously used AI for non-generative purposes, including automatic captioning, which underscores the distinction the company is now drawing between assistive AI features and generative model training.
What happens next?
Twitch's help documentation indicates that the opt-out setting can be changed immediately within account settings, meaning creators can act on the choice right away rather than waiting for a rollout period, according to the same technology publication. The company has not announced a timeline for ending default data collection altogether, nor has it committed to publishing an audit detailing what user content may already have been used for training, based on available reporting.
Amazon has been approached for comment.
Key Facts
- Amazon bought Twitch in 2014 for close to $1bn (£740m).
- The opt-out setting is found under Security and Privacy, labelled 'Training for Generative AI'.
- It covers streams, VODs, clips, chats, and channel images and text, but not captioning or safety tools.
- Twitch product chief Mike Minton said data collected would not be resold to third parties.
- No date has been set for ending default collection or publishing a usage audit.
