A German Court Says an AI Comic Made From a Dog Photo Doesn't Infringe Copyright

A German appeals court has ruled that turning a copyrighted photograph into an AI-generated comic-style image does not automatically infringe the photographer's rights. The decision came out of a dispute over an underwater shot of a dog reaching for a red toy.

According to the suit, the photographer sells underwater dog portraits and sued a former cooperation partner, the operator of a dog school, who uploaded one of her photos into AI software, prompted it to generate a new image, and published the result on his website. When a cease and desist letter went nowhere, she applied for a preliminary injunction.

The court involved was the Düsseldorf Higher Regional Court, and its decision on April 2, 2026 came in preliminary injunction proceedings, not a full trial on the merits. The Düsseldorf Higher Regional Court affirmed the lower court's decision, though it arrived there by a different line of argument. The lower court had treated the AI image as a permitted "free adaptation." The appeals court disagreed with that framing, then still found no direct infringement.

The original photo showed the dog underwater with tight framing and shallow depth of field, and the court noted that practically only the dog's head and the toy were visible because the body almost disappeared through the chosen perspective and blur. The court regarded that interplay of sharpness and movement as the artistic accomplishment. The AI image kept the same idea but rendered the whole dog in a flat cartoon style with exaggerated paws. As legal analysts put it, a dog reaching underwater for a toy is, as a concept, not eligible for protection; the photographer came across the scene rather than creating it. The court concluded the AI output had transformed the protected elements: framing, angle, lighting, and sharpness. Being able to recognize the same dog and setting was not enough on its own to show copying.

It helps to place this next to the other German AI copyright fight involving a photographer. Stock shooter Robert Kneschke sued the dataset operator LAION after finding his images in its collection, and the Higher Regional Court of Hamburg dismissed his appeal, confirming that creating the LAION-5B dataset did not infringe copyright. That case was about training data, so it is a different question from this one, which is about the output. Read together, the two point in the same uncomfortable direction for working photographers: German courts have so far declined to stretch existing copyright to cover either the ingestion of images for training or AI outputs that borrow a subject without copying the specific creative choices. Both rulings still have room to move, and the Kneschke case was cleared for a further appeal to Germany's Federal Court of Justice.

The practical takeaway is not "AI can freely copy my work." It is that copyright, in Germany and across the EU, protects your specific expression rather than the thing you pointed the camera at. A rival can shoot the same dog with the same toy, and now the argument is that an AI can be prompted toward the same motif, as long as it does not carry over your framing, light, and depth of field. That is a narrower shield than many shooters assume they have. If your business depends on a repeatable concept, a distinctive underwater look, a signature setup, a recognizable series, this is a reason to think about contracts, licensing terms, and who has access to your files, because a former partner with your originals is exactly how this case started. These are lower-court decisions in one country, not a settled global rule, and they expose gaps and inconsistencies in how existing copyright law handles AI rather than closing the question. The next ruling could sharpen the line in a different place.