New Twitch Feature Uses Channel Content to Train AI Without Asking Permission

Twitch revealed that it is utilizing streamer content for Amazon's AI model training purposes without requiring permission. The feature, which was reported on Wednesday, is now accompanied by an opt-out option in the platform's settings to give creators the ability to disable the feature. However, it is enabled by default unless users take action to turn it off.

Amazon's AI models, like those used for Alexa Plus and other projects, necessitate vast amounts of data to train and enhance language or image accuracy. The streaming service is teeming with content such as videos, audio, and chat transcripts, which can be used for training. Nonetheless, individuals who have not given consent may be uncomfortable with their voice, words, or video being analyzed and utilized in future AI prompts.

Amazon and Twitch representatives did not provide immediate responses to queries regarding the matter. According to Twitch's latest FAQ, when users permit training, their content might be used to enhance upcoming Generative AI models. For instance, audio from users might aid in refining speech-to-text models, benefiting Twitch's captions and Amazon's captions as well.

Users can still utilize other AI-powered features like Automods and campaign optimization even if they opt out of training. The impact on chat during a stream remains unclear, as Twitch has not mentioned chat or its relationship with the training feature. It is essential to consider consent issues surrounding audience members' contributions and the stream itself, as Twitch has not provided information about this aspect.

Many Twitch users also stream other live content, such as other creators' channels, which may create a conflict when the original streamer opts out of AI training, but the streamer rebroadcasting their content has AI training enabled. In such situations, the AI model may still access the original streamer's content without their explicit consent.

Additionally, Twitch has faced legal challenges concerning copyrighted music, and some content creators have found ways to bypass the platform's music restrictions. If the training data collection method is well-defined, Amazon could potentially gather copyrighted music from streaming services for use in its AI models, leading to additional legal concerns.