Twitch allegedly used creator content to train Amazon AI, a new class action lawsuit alleges

On Aug. 20, 2026, Twitch creator Warren Pandiscia filed a class action lawsuit against Twitch Interactive Inc. and parent company Amazon.com Inc. in the U.S. District Court for the Northern District of California.

The lawsuit alleges the companies enrolled every Twitch channel in an artificial intelligence training program by default then fed years of creator streams, videos, clips, chat logs and channel page text into Amazon's generative AI products without permission.

Pandiscia's experience

Pandiscia reportedly streamed Lego builds and video games on Twitch for close to 10 years, logging more than 1,000 hours and drawing more than 900 followers. He claims he turned the AI training setting off as soon as he learned it existed although the filing alleges Amazon already used streamers' content before providing the opt-out option.

Pandiscia seeks to represent a nationwide class that includes every creator whose content Amazon used to train its AI without consent. Between 3.2 million and 6.9 million creators go live on Twitch each month.

A default setting

Twitch reportedly activated a setting labeled "training for generative AI" across every account on Aug. 12, 2026. There was no email, pop-up or announcement preceding the launch, the lawsuit claims. A reporter allegedly found the toggle.

Chief product officer Mike Minton defended the default during a live Q&A, saying, "If it was opt-in, nobody would opt in. That's honestly the answer."

Some creators allege the toggle turned back on after they switched it off, and because the setting attaches to a channel rather than a person, chat messages fall under whoever owns the channel, according to the lawsuit.

The program predates the toggle, the complaint contends. Minton, then chief monetization officer, said at a 2024 event that Amazon already used Twitch content to build AI, though only in early testing and not at full scale.

Did Twitch's old terms allow AI training?

The complaint claims Twitch rewrote its terms and conditions documents the same day the setting went live to allow AI training.

The prior terms of service licensed creator content "in connection with monetizing the Twitch Services." The amended version includes "Twitch and/or its affiliates' business," wording the filing calls broad enough to cover Amazon's AI operations. The privacy notice also reportedly named AI training as a purpose for the first time. None of those edits would have been necessary if the old terms already allowed the training, the suit reasons.

The proposed class action claims Twitch broke two privacy notice promises: not to weaken protections on information it already collected without asking users first and to give notice and a chance to decline before sharing data with third parties.

The legal claims

The lawsuit brings four claims:

  • Breach of implied contract and the implied covenant of good faith and fair dealing, against Twitch over expectations that years of platform conduct created
  • Unjust enrichment, against both companies, which bars a business from keeping value it gained unfairly
  • Breach of express contract, based on the written terms of service and privacy notice
  • California's Unfair Competition Law, which the suit pleads under all three prongs, unlawful, unfair and fraudulent, with the state wiretapping law as one predicate

What the case means for Twitch creators

The proposed class action seeks damages, restitution, disgorgement of profits and an injunction ordering both companies to halt unauthorized training, implement a durable opt-out and delete the information they took without consent.

Creators who want to opt out can disable the setting under the security and privacy section of their account settings. As of this writing there is no settlement, no claims process and no money available.