Twitch is facing a class-action lawsuit over allegations that Amazon used content from the streaming platform to train its generative artificial intelligence models. The suit, filed in California, claims that Amazon accessed Twitch streams and videos to create a massive dataset for its AI products without obtaining proper licenses or seeking permission from content creators.
The plaintiff, Warren Pandiscia, a streamer with approximately 1,000 followers, stated in the complaint that Amazon had a significant incentive to acquire training data on an unprecedented scale due to the commercial nature of its AI products. The lawsuit argues that these actions were not only unlawful but also constituted an "unconscionable attack on the community of content creators" whose work fuels the generative AI industry without any compensation.
According to the complaint, Twitch's system for AI training was an opt-out feature, a design choice that Twitch's chief product officer reportedly admitted was implemented because "nobody would opt in." The suit further contends that the consent mechanisms are inadequate, as a streamer participating in another person's stream could have their voice used for AI training depending on the host streamer's settings, regardless of their own opt-out status. The legal filing emphasizes that Amazon's systems are incapable of obtaining consent from all parties involved in captured communications.
Furthermore, the lawsuit accuses Twitch and Amazon of failing to provide adequate notice about the data collection feature. The complaint states that "Twitch sent no email, displayed no pop-up notification, and made no announcement" before enabling the setting for all accounts. The feature was reportedly discovered by a reporter, Zach Bussey, rather than being proactively communicated to users.