Stability AI Sued by Getty Images and Artists for Training Stable Diffusion on Copyrighted Images

In January 2023, Getty Images filed a lawsuit in Delaware federal court against Stability AI, alleging that the company unlawfully copied and processed millions of images from Getty's collection to train its Stable Diffusion AI image generator. The lawsuit claimed that Stability AI scraped Getty's copyrighted images without permission as part of training the model on the LAION-5B dataset, which contains approximately 5.85 billion image-text pairs collected from across the internet. Separately, a class-action lawsuit was filed by artists Sarah Andersen, Kelly McKernan, and Karla Ortiz against Stability AI, Midjourney, and DeviantArt in the Northern District of California. The artists alleged that these companies violated copyright law by training their AI models on billions of copyrighted images without consent or compensation. The complaint argued that the AI systems create derivative works that compete directly with the original artists whose work was used in training. The legal challenges center on whether training AI models on copyrighted material constitutes fair use under copyright law. Getty Images claimed that Stability AI's use was commercial in nature and harmed the market for licensed images. The company pointed to instances where Stable Diffusion generated images containing corrupted versions of Getty's watermark as evidence of direct copying. Artists in the class-action suit argued that the AI systems learned to replicate their distinctive artistic styles, effectively creating unauthorized derivatives. Stability AI defended its practices by arguing that training AI models on publicly available images constitutes fair use, similar to how humans learn by observing existing art. The company contended that the AI does not store or reproduce copyrighted images but rather learns statistical patterns to generate new, original works. However, critics pointed out that the AI can be prompted to create images "in the style of" specific artists, potentially undermining their market value. The lawsuits seek monetary damages, injunctive relief to prevent further alleged copyright infringement, and destruction of AI models trained on copyrighted works. The cases are ongoing and could establish important precedents for how copyright law applies to AI training data. Similar lawsuits have been filed in the UK, where Getty Images is also pursuing legal action against Stability AI. The broader implications extend beyond Stability AI to the entire generative AI industry, as most large-scale AI models rely on training datasets that include copyrighted material scraped from the internet. The outcome could significantly impact how AI companies source training data and potentially require fundamental changes to current industry practices around dataset creation and model training.