Twitch Sued Over Amazon AI Training Setting
The case hinges on whether a default-enabled AI training setting, disclosed only after implementation, breaches Twitch's contractual promises to creators, a question that could set precedent for how platforms obtain consent for generative AI training data.
Key Facts
- Warren Pandiscia filed a class action lawsuit against Twitch and Amazon on August 20 in the U.S. District Court for the Northern District of California.
- Twitch added the "Generative AI Training" setting on August 13, enabled by default, allowing Amazon to use channel content to train generative AI models.
- Twitch Chief Product Officer Mike Minton said the setting was opt-out because "if it were opt-in, nobody would opt in."
- The lawsuit requests class certification, injunctive relief, disgorgement of profits, damages, and litigation costs.
- Even users who opt out may still have their content used if it appears on channels that have not opted out.
Reporting from 3 sources: Automaton, GameBusiness.jp, GIGAZINE.
Twitch content creator Warren Pandiscia filed a class action lawsuit on August 20 in the U.S. District Court for the Northern District of California against Twitch and its parent company Amazon. The suit targets a new "Generative AI Training" setting added on August 13 under the Security and Privacy tab. This setting, enabled by default, lets Amazon use channel content to train generative AI models, including those that convert audio to text and improve captions across Amazon services. Pandiscia argues that Twitch and Amazon used streamer content for AI training, development, and commercialization without adequate notice or consent, exceeding what Twitch's Privacy Notice allowed. He also claims the companies violated the good faith principle and unfairly profited from user content. The complaint requests class certification, appointment of Pandiscia as class representative, a declaration that the actions were unlawful, disgorgement of profits, damages, and injunctive relief to stop AI training. Twitch Chief Product Officer Mike Minton said the setting was opt-out because an opt-in approach would result in no one permitting use. Even users who opt out may still have their content used if displayed on channels that have not opted out.
Twitch Chief Product Officer Mike Minton's full remarks during the August 12 patch notes stream went further than the summary states. Asked by viewers why the AI training consent setting defaults to allowed, Minton replied, "Honestly, if it were opt-in, nobody would opt in." He added that "what we're doing here that's unique and different is respecting people's wishes to opt out of AI training," then acknowledged, "I know this isn't something fans will accept. I understand it's very disappointing for the community, but this is the current situation." The candid answers drew criticism from streamers who expected a dodged question. Pandiscia told GIGAZINE that "reading the terms of service, there was no way to anticipate such a possibility," and he would have opted out earlier if the use had been disclosed in advance. His complaint describes streamers as "free training material for another commercial AI product" and alleges they "lost money or property as a result of defendants' illegal, unfair, and fraudulent conduct." The August 13 disclosure came via Twitch's official X account, which stated that "across Amazon, your channel's content will be used to train generative AI content models." This is not the first AI training suit involving creator content. The same three YouTube channels that sued Apple in April over millions of scraped videos also filed against Meta, NVIDIA, ByteDance, and Snap.
Synthesized by Yomimono from the 3 cited sources below, including Japanese-language reporting where cited, then editorially reviewed before publishing.