Moscow court rules AI-generated images are technical process, not artwork

The plaintiff fed images of Jukonda and the Statue of Liberty into a neural network, swapped the torch for a wine glass, shaved the hair and added bags under the eyes, and claimed the resulting pictures as his own creation. A second entrepreneur later downloaded the finished files without permission, printed them on T-shirts, hoodies and thermos mugs, and sold them on his website. The plaintiff sued for compensation.

The judges rejected the claim, describing the work with the neural network as a technical procedure rather than a creative act. They noted that the idea of swapping the torch for a wine glass is not new and has been used in art for decades, concluding that both the prompt and the composition fail to meet the originality threshold required under Russian law. One of the judges labeled the concept “unoriginal idea.”

The ruling draws a clear line: operating a generative model constitutes a technical action, not an expression of personal authorship, but a statistical calculation of model weights. The court did not address whether the model itself infringes copyright during training; it simply held that the final output does not belong to the operator. The decision was summed up as “technical, not creative.”

As a result, anyone in Moscow who sells merchandise featuring images generated by Midjourney or Stable Diffusion will have no legal protection. The images are considered public domain the moment they leave the model, a stance opposite to the United States, where the Copyright Office refuses registration of pure AI-generated pictures but allows registration when there is substantial human editing. Russia’s decision closes that avenue entirely.