Content creators on the live-streaming platform Twitch are facing a new challenge with the use of their data to train artificial intelligence. It has been revealed that a setting allowing the use of channel content in training Amazon's generative AI models is enabled by default and can be manually disabled by users. The issue is particularly sensitive as Twitch channels contain not only short texts but also hours of video, audio, and conversations produced by creators and their audiences.

A switch that needs to be turned on
 

According to a report published on Engadget, Twitch, which is owned by Amazon, has added an option within the settings that allows channel owners to opt-out of using their content in training generative AI models. This includes live streams, recorded copies, clips, channel conversations, text, and images. However, the controversial point is that the setting was enabled by default, meaning users had to take action themselves if they did not want their content used for this purpose.

Withdrawal does not stop all uses of artificial intelligence
 

Twitch clarifies that disabling generative training does not mean stopping all uses of artificial intelligence or machine learning within the platform. Some systems, such as automated translation tools, moderation, and community protection, remain separate from this option. An additional issue arises in conversations, where viewers' written content usage can be affected by the settings of the channel owner they are participating in. This brings back an older debate with more urgency: Should training models on user-generated content be a voluntary choice or an automatic setting that requires finding a way to disable it?