Twitch is officially facing severe legal trouble from its own streamers. On August 20, a Connecticut-based content creator filed a class-action lawsuit against Twitch and its parent company, Amazon. The plaintiff, Warren Pandiscia, alleges that Twitch unlawfully harvested creator data for Amazon’s generative artificial intelligence (AI) training programs. Consequently, this lawsuit represents a proposed class of all Twitch video streamers who are demanding compensation for their misused intellectual property for AI training.
Twitch AI Training Lawsuit: Years of Unpaid Data Harvesting
According to the 37-page complaint filed in the Northern District of California, Amazon and Twitch copied millions of videos without securing permission. Pandiscia claims this massive data collection dates back to as early as 2024. During that year, Mike Minton, who was Twitch’s Chief Monetization Officer at the time, admitted the company used content for AI “prototyping”. Back then, Minton insisted the practice stayed within privacy regulations. However, the platform never directly notified streamers about this underlying data scraping.
Because Amazon commercializes its AI products, the lawsuit argues the tech giant had an overwhelming incentive to acquire massive datasets. Instead of negotiating lawful licenses or paying creators, Amazon simply took the data. The scraped content reportedly includes past broadcasts, live streams, clips, chats, pictures, text, and audio. Ultimately, Amazon uses this data to refine models that synthesize text, audio, images, and video, such as improving speech-to-text captions across Amazon services.
The Controversial Opt-Out Policy
The situation rapidly escalated earlier this month. On August 12, Twitch published a post on X confirming it uses channel content to train generative AI models. Simultaneously, the platform updated its Terms of Service and Privacy Policy. Twitch introduced a new setting that lets streamers deny Amazon access to their data. However, the company automatically opted every single channel in.
Streamers quickly expressed outrage over the default opt-in status. In response, Minton, now serving as Chief Product Officer, defended the decision during an August 12 stream. He bluntly stated:
If it was opt-in, nobody would opt-in. That’s honestly the answer.
Therefore, rather than respecting user privacy from the start, Twitch intentionally forced creators to manually opt out to secure its necessary training data.
System Flaws & Legal Demands
Unfortunately, simply opting out does not fully protect creators. The lawsuit highlights a major structural flaw: the privacy setting only applies per channel. If an opted-out streamer appears in the chat or broadcast of an opted-in channel, Amazon still harvests their data. Furthermore, Pandiscia notes that Twitch’s systems are completely incapable of obtaining consent from all parties involved in these captured communications. Moreover, creators will never be able to claw back the two years of historical data that Amazon has already consumed.
Pandiscia asserts he never gave permission for this usage and opted out immediately after discovering the buried setting. He claims these actions actively cost him money, property, and control over his original work. As a result, the lawsuit accuses Twitch and Amazon of unjust enrichment, unfair business practices, and breaching both implied and express contracts. The plaintiff now asks the court for injunctive relief, damages, restitution, and a disgorgement of profits from Amazon, which originally acquired Twitch for roughly $970 million in August 2014.
