Warner Bros. alleges that AI models like Midjourney profit from generating images of iconic characters such as Scooby-Doo, Superman, Wonder Woman, and Batman, trained without consent, according to Guides Csbsju. This claim exposes a core conflict in the visual arts, where AI’s rapid development challenges established intellectual property rights.

AI art models democratize image creation and increase market supply, but they simultaneously undermine the livelihoods and intellectual property rights of the human artists whose work they were trained on. This creates a profound cultural impact, forcing the 2026 art market to navigate technological advancement and ethical boundaries.

Based on the rapid market shifts and ongoing legal battles, the art world is likely heading towards a bifurcated future where human art is either highly niche and valued for its provenance, or completely overshadowed by AI, unless robust regulatory frameworks are established.

The Legal Battleground: Unauthorized Training and Copyright Claims

Artists Sarah Andersen, Kelly McKernan, and Karla Ortiz filed a class-action lawsuit against Stability AI, DeviantArt, and Midjourney, challenging Stable Diffusion's use, according to Guides Csbsju. This lawsuit directly questions the core practice of training AI models on massive datasets without explicit creator consent.

The New York Times also sued OpenAI in 2024 for copyright infringement, alleging unauthorized content use for training algorithms, according to Gsb Stanford. These diverse legal actions, from individual artists to major media corporations, reveal a widespread challenge to AI companies' data sourcing methods, implying that the entire training paradigm faces an existential threat.

The LAION-5B image set, used to train AI art models with 5.8 billion images, was later found to contain personal medical record photos, according to Research Guides. Stability AI's decision to allow artists to remove their work from the Stable Diffusion 3.0 training dataset comes only after billions of images, including those with privacy concerns, were already incorporated into existing models. This belated opt-out system highlights the retroactive nature of ethical considerations, suggesting that privacy and consent were secondary to rapid data acquisition.