A Connecticut streamer says Twitch and Amazon fed his broadcasts into AI models for two years without asking, then quietly built an opt-out button nobody was told to look for.
Warren Pandiscia streams to more than 900 followers on Twitch. That's not Twitch royalty. But on August 20, he became the lead plaintiff in a class-action lawsuit filed in the Northern District of California, accusing Amazon and Twitch of scraping millions of hours of streamer video since at least 2024 to train generative AI models, without consent and without paying anyone a cent.
The complaint doesn't mince words. It argues Amazon needed enormous amounts of training data fast. Rather than license it or ask permission, it simply took what was already sitting on Twitch's servers. Pandiscia is pursuing claims of breach of implied and express contract, unjust enrichment, and unfair business practices. Once footage trains a model, you can't get it back. That's the lawsuit's core point. "Content creators such as plaintiff and the class members will never be able to claw back the intellectual property unlawfully copied and used by defendants to train Amazon's generative AI," the complaint states, as Courthouse News Service reported.
An Opt-Out Nobody Was Told to Find #
Here's the thing that makes this case land differently than a typical data-scraping dispute. On August 12, Twitch rolled out a setting letting streamers block their content from AI training. It covers streams, VODs, clips, stream chats, and even the images and text on a channel page. Sounds like a fix. It isn't one. The setting is switched on by default. Every creator was automatically enrolled in AI training the moment it launched, and had to go find the toggle themselves to get out.
Twitch's own chief product officer, Mike Minton, said the quiet part out loud during a livestream addressing the backlash. "If this was opt-in, nobody would opt in. That's honestly the answer," he said, according to Forbes. That's a remarkable thing for a company executive to admit on the record. It confirms what creators already suspected: the default wasn't an oversight, it was the strategy. Make consent the thing users have to actively escape, and most of them never will.
Amazon Quietly Raised Prices on Echo, Kindle, Fire TV and Eero Devices Amazon quietly raised prices overnight on Echo, Kindle, Fire TV and eero devices, with the Echo Dot jumping 60% to $79.99. According to Fortune, an Amazon spokeswoman confirmed the industry is facing steep memory and storage cost increases tied to AI demand, while Ring products were left untouched. - amazon device price increases memory chip shortage - why did amazon raise echo and kindle prices
You don't need a law degree to see the tension. Twitch's own terms of service govern how creator content can be used, and the lawsuit alleges the AI training program breached those terms outright. If a platform needs to bury a toggle inside settings menus to get the participation numbers it wants, that's a pretty strong signal the practice wouldn't survive an honest ask.
Not an Isolated Case #
Amazon has leaned hard into generative AI investment over the past two years. Twitch, which it bought in 2014 for about $970 million, sits on a mountain of exactly the kind of raw video and speech data that trains multimodal models well. Millions of hours of people talking, reacting, gaming, and interacting with live chat is a goldmine for anyone building conversational or video AI. That's the appeal. The lawsuit says Amazon should have paid for it, or asked first.
This case also isn't happening in isolation. Getty Images, The New York Times, and a long list of authors and artists have all sued AI developers over training data pulled from copyrighted or user-generated work without permission. What makes the Twitch suit distinct is the platform relationship. Streamers didn't publish their videos on some open corner of the internet expecting a crawler to find them. They uploaded to a platform they had a direct account and terms-of-service relationship with, one that then allegedly used that relationship to harvest their work internally.
Opt-out-by-default is spreading well beyond streaming. Plenty of software platforms now bury AI-training permissions in settings pages most users never open, betting that inertia does the consenting for them. If a court finds that Twitch's version of this crosses a legal line, breaching its own terms of service in the process, that ruling won't just apply to gaming. It becomes a template other platforms have to think twice about copying.
Amazon and Twitch haven't filed a public response to the complaint yet. Pandiscia is seeking class-action status on behalf of every streamer whose content was used the same way. Given Twitch's user base, that could run into the millions.
Also read: OpenAI's Chris Lehane warns AI hacking is turning into a permanent threat • How One Judge's Split Ruling on Anthropic Became AI's Copyright Rulebook • Flock Safety's CEO Calls For Compromise As Cameras Draw Fire From Both Sides