Streamers File Class-Action Lawsuit Against Twitch and Amazon Over AI Training on Their Streams

The Twitch platform and its parent company Amazon have faced a class-action lawsuit alleging that they used streamers' content and video recordings from the platform to train generative artificial intelligence models. According to the plaintiffs, the tech giant collected and used creators' broadcasts without prior permission, without any lawful licensing agreements, and without any form of compensation. The case was triggered by a lawsuit filed by content creator Warren Pandishia, who is acting as the representative for a group of authors consolidated into a class action.

The Core of the Claims: Content Collected Without a License

The court documents state that the mass collection and use of streamers' broadcasts for machine-learning purposes began as early as 2024. A key element of the complaint is the absence of explicit consent from the authors: according to the plaintiffs, no licensing agreements or permission requests were ever executed, and the content was taken into use by default. This means that, in the authors' view, their creative material was effectively used as raw feedstock for training models without their knowledge and without any financial participation in the results.

The Opt-Out Mechanic and Twitch's Public Stance

The situation intensified against the backdrop of the platform's introduction of an opt-out feature, which allows authors to manually prohibit the use of their channels for training purposes. Twitch's Chief Product Officer Mike Minton previously openly defended precisely this mechanic, stating that under an alternative opt-in format—where consent must be given explicitly—no one would voluntarily agree. In effect, the company has locked in a model of "silent consent by default with the right to opt out," which the plaintiffs consider legally and ethically unacceptable.

Contradictory Accounts

Here, two versions clash directly. The Twitch and Amazon side regards opt-out as a sufficient and reasonable mechanism: the company believes it has provided authors with a tool of control, and that a shift to opt-in would make content use impossible due to mass refusal. The plaintiffs, by contrast, argue that opt-out is not consent: in their version, lawful use requires precisely active, informed permission (opt-in), and the "silent" inclusion of content in training since 2024 without licenses or compensation is a violation of authors' rights. The difference between the two versions comes down to what exactly counts as an acceptable basis for using someone else's content in machine learning, and it is on this very question that the court's argumentation will be built.

What the Authors Are Seeking and the Possible Consequences

The plaintiffs are asking the court to issue an injunction ordering the cessation of the use of their content, as well as to recover compensation for the damages suffered and to return the illegally obtained benefit. If the court sides with the authors, this could force Twitch and Amazon to rethink their entire model of handling user content in the context of AI—from the wording of the terms of service to the very logic of data collection for training. For the broader streaming and generative-AI industry, the outcome of the case will become an important precedent: it will determine whether a "right to opt out" is enough, or whether using someone else's broadcasts in machine learning always requires explicit prior consent.