Twitch Accused of Using Livestreams for Unconsented AI Training

Twitch has been accused of using livestreams to train Amazon's generative AI without creators' consent. A lawsuit filed in California on August 20 claims that Twitch rewrote its Terms of Service (ToS) on August 12, allowing the platform to use content for AI training without explicit permission.

The complaint cites Section 8(a)(i) of Twitch's ToS, which was amended on August 12. The original language granted Twitch a license tied to monetization, while the revised version allows the company to use content 'in connection with Twitch and/or its affiliates' business', effectively allowing Amazon to train AI models using user content.

The lawsuit also points out that the amended ToS added a new recital stating that users are directing Twitch to cache and store their User Content, including individual pieces of it. This change is seen as evidence that earlier terms did not convey the rights defendants exercised.

Twitch's chief product officer, Mike Minton, was quoted defending the platform's opt-out setting, saying 'If it was opt-in, nobody would opt in.' However, when asked whether Amazon uses Twitch to train AI models, he replied, 'Yeah, for sure', but added that it was only for prototyping purposes.