Twitch Confirms User Content Trained Amazon AI, Adds Opt-Out

Amazon-owned streaming platform Twitch has acknowledged that user-generated content, including streams and other material, has been used for years to train the company’s generative AI models, according to a report from Ars Technica.

Ars Technica reported that Twitch’s terms note user content “may be used for future Gen AI model improvements,” confirming that such training has already been taking place. The platform is now giving users the ability to opt out of having their content used for this purpose going forward, per the report.

The disclosure adds Twitch to a growing list of major platforms facing scrutiny over how user-generated content is repurposed to train artificial intelligence systems without explicit upfront consent. As Ars Technica noted, the practice raises questions common across the tech industry about data rights, transparency, and the extent to which platforms can use material created by their users for commercial AI development.

Details on the scope of the training, including which specific Twitch content was used and over what time period, were not fully specified in the available reporting. It also remains unclear what mechanisms Twitch has put in place to verify opt-out requests or whether the opt-out applies retroactively to content already used in training existing models.

The move comes amid broader legal and regulatory attention on AI training data practices, as courts and lawmakers weigh how copyright law applies to the use of user-generated and copyrighted material in building generative AI systems. Twitch’s disclosure, as reported by Ars Technica, illustrates how platforms tied to major tech companies, in this case Amazon, are adjusting policies in response to mounting public and legal pressure over AI training transparency.

Based on reporting by arstechnica.com.