DeviantArt DreamUp AI Trained on Artist Works Without Consent

On November 9, 2022, DeviantArt announced the launch of DreamUp, an AI-powered image generator integrated into their platform and based on the Stable Diffusion model. The announcement immediately triggered fierce backlash from the artist community when it became clear that the underlying AI model had been trained on millions of artworks from DeviantArt and other platforms without explicit artist consent. The Stable Diffusion model used LAION-5B dataset containing over 5 billion image-text pairs scraped from across the internet, including substantial portions of DeviantArt's catalog spanning two decades of user-generated content. The controversy intensified as artists discovered they could generate images in specific artists' styles by simply prompting the AI with artist names, effectively allowing users to create derivative works that mimicked established creators' distinctive techniques and aesthetics. High-profile artists including Greg Rutkowski, whose name became one of the most popular prompts, found their artistic styles being replicated without permission or compensation. The situation was particularly problematic given DeviantArt's position as a trusted platform where artists had uploaded their works expecting protection of their intellectual property rights. Within hours of the announcement, thousands of artists began protesting through social media campaigns, petition drives, and threats to leave the platform entirely. A Change.org petition titled 'DeviantArt: Stop AI Art Generation Unless Explicit Opt-in Consent Given' rapidly gathered over 10,000 signatures. Artists argued that using their works to train commercial AI systems without consent violated both copyright law and basic ethical principles of creative attribution. Many expressed concerns that AI-generated art would devalue human creativity and potentially eliminate career opportunities for emerging artists. Faced with overwhelming community opposition and potential legal challenges, DeviantArt initially attempted damage control by implementing an opt-out system called 'noai' tags and later DreamUp's opt-out feature. However, this retroactive approach was criticized as insufficient since the training had already occurred and the damage was done. The company also faced technical challenges in implementing effective opt-out mechanisms, as removing specific works from already-trained models proved practically impossible. The incident highlighted fundamental tensions between technological innovation and artist rights, contributing to broader legal challenges against AI companies and ongoing debates about fair use in machine learning training datasets.