Twitch Streamer Sues Amazon, Twitch Over 1,000 Hours of AI Training Content

3 articles · Updated · TNW · Aug 24

A 37-page complaint filed in San Francisco on Aug. 20 says Twitch used Warren Pandiscia’s streams to train Amazon generative AI without consent, seeking class status for creators whose content was allegedly used.
The suit centers on Twitch’s Aug. 12 policy rewrite, which changed content rights from monetizing Twitch services to affiliates’ business, added AI language to its privacy notice, and launched a default-on opt-out setting.
Pandiscia argues the earlier terms cut off Twitch’s license when content was deleted and barred data-mining or automated extraction, making the later amendments evidence that prior contracts did not authorize AI training.
Mike Minton is quoted saying an opt-in system would fail because “nobody would opt in,” and separately saying he did not know what Amazon had already used for model training.
The proposed class could cover millions of creators and seeks damages, disgorgement, and an order to identify, segregate, and delete unlawfully obtained training data and any models derived from it.