Source: Aftermath
Amazon is converting Twitch's massive archive of unstructured video—millions of hours of gameplay, commentary, and ambient content—into raw material for training multimodal AI models, effectively monetizing creator output without explicit consent or compensation beyond platform access. This differs from licensing deals like YouTube's with AI companies: Twitch is extracting data unilaterally rather than negotiating rights. The precedent is direct: livestreaming platforms with vast video catalogs can become de facto training data mines, while creators retain minimal control over downstream commercial use of their work.