Twitch streamer sues Amazon over AI training on his content

Warren Pandiscia, a Twitch streamer based in Connecticut with more than 900 followers, filed a class action lawsuit against Twitch and its parent company Amazon on August 20, 2026, in the U.S. District Court for the Northern District of California. His 37-page complaint, first reported by Courthouse News Service, accuses the two companies of copying millions of streamers' videos without permission or compensation to build the datasets that train Amazon's generative artificial intelligence products.

A dataset built without asking

"Because Amazon AI products are commercialized, Amazon had an overwhelming incentive to acquire training data on an unprecedented scale. Rather than negotiate for lawful licenses or seek permission, defendants accessed the Twitch streams and videos to utilize them as a massive dataset necessary to fuel Amazon's AI products," Pandiscia wrote in the complaint, according to Courthouse News. He claims Twitch and Amazon began scraping streamers' content for this purpose as early as 2024, well before either company said anything publicly about it.

Pandiscia says he had "no reason to believe" his content would be used to train generative AI products and is "opposed to such use of his original content." He says he opted out of the training setting as soon as he learned it existed, and would have done so earlier had Twitch disclosed the practice sooner.

"If it was opt-in, nobody would opt in"

On August 12, 2026, eight days before the lawsuit was filed, Twitch Support announced on X that it had added a setting letting streamers opt out of "having your channel content used to train generative AI content models across Amazon," linking to an FAQ page. The same day, Twitch updated its Terms of Service and Privacy Policy with new language covering automated tools and artificial intelligence. According to the FAQ, Amazon may use streams, past broadcasts, clips, chats, pictures and text for AI training, including future models built by Amazon to generate or synthesize text, audio, images or video.

Crucially, the setting is enabled by default: every Twitch channel is automatically opted in to AI training unless the streamer manually opts out. Settings also apply by channel rather than by individual, so content from a viewer who opted out can still be captured if it appears on a channel that remains opted in.

If it was opt-in, nobody would opt in.

Minton gave that answer shortly after the opt-out setting launched, when a viewer asked during Twitch's recurring Patch Notes stream why AI training wasn't opt-in to begin with. Pandiscia's complaint argues that a system in which the only way for a user to give informed consent is to accidentally learn a setting exists is not really a consent system at all: "By design, defendants never obtain—and their systems are incapable of obtaining—the consent of all parties to the communications they capture," the complaint states, according to Courthouse News.

The legal claims and what Pandiscia is seeking

• Breach of implied contract: Twitch allegedly used streamers as "free training stock for separate commercial AI products" without the licensing users believed governed their content.
• Breach of express contract: the complaint says Twitch broke earlier commitments to obtain consent before making its policies less protective and to notify users before sharing their data with third parties.
• Unjust enrichment: Amazon and Twitch are accused of profiting from streamers who "lost money or property" as a result of the alleged practices.
• Unfair business practices under California law, tied to the same underlying conduct.

Pandiscia is asking the court for injunctive relief, along with damages, restitution and disgorgement of the profits Amazon and Twitch allegedly earned from streamers' content. Amazon acquired Twitch in August 2014 for approximately 970 million dollars, a deal that made the streaming platform a wholly owned subsidiary rather than an independent company able to set its own data policy.

The case adds Amazon to a growing list of technology companies facing litigation over how they source AI training data from creators. Earlier in 2026, three YouTubers filed a separate class action against Apple, accusing the company of scraping copyrighted video content to train its AI models without authorization.

What this means for creators everywhere

For streamers and creators outside the United States, the case is a reminder that the platforms hosting their work can also be training grounds for AI systems they never agreed to supply. Default-on settings, buried in account menus and explained only in FAQ pages, are becoming the industry's standard method for securing training data at scale — and Minton's own words suggest platforms know an opt-in model would starve their pipelines. Whatever the outcome in San Francisco, the underlying question — who owns the value in a creator's back catalog once a platform's parent company starts building AI products from it — is not confined to one country's courts, and any streamer whose content sits on a major platform faces the same default-on choice Pandiscia says he never meaningfully made.