Amazon and Twitch Sued Over AI Training Data

Amazon and Twitch are facing a class-action lawsuit alleging that the companies used videos broadcast on the streaming platform to train artificial intelligence models without the creators' permission or proper compensation. The lawsuit was filed on behalf of millions of Twitch streamers whose content was allegedly used to develop Amazon's AI.

Connecticut-based streamer Warren Pandiscia is the lead plaintiff in the suit, which claims that Twitch breached its user contract by employing the content as AI training data for Amazon's models. The suit argues that streamers' work should not be commercially exploited for AI development without their consent or remuneration.

The lawsuit seeks damages for streamers whose content was purportedly used for AI training and an injunction to halt the alleged practice. It remains unclear when Amazon began gathering Twitch user data for AI model training. In August, the company announced the move, offering users an opt-out option for the use of their data.

Twitch's chief product officer, Mike Minton, has previously stated he was unsure if user data had been scraped for training purposes before the opt-out was introduced. He also expressed uncertainty about precisely what Amazon had used for its model training.

Twitch's own frequently asked questions regarding AI training state that user audio might be utilized to refine speech-to-text models, thereby improving automatic subtitles on both Twitch and Amazon videos. Streamers can disable this feature by navigating to their Streamer Dashboard settings, selecting Security and Privacy, and toggling off "training for Generative AI." However, this opt-out applies only to individual streams, meaning content from creators who appear in streams where the feature remains enabled could still be used for AI training.

Amazon, which acquired Twitch in 2014 for nearly $1 billion, has identified AI as a significant focus area as it competes with major tech companies like Google and Meta. Streamers on the platform reportedly produced over 215 million hours of content in the first few months of 2026 alone, highlighting the vast amount of data available for potential AI training.