German AI Image Copyright Ruling: Why the Dog Photographer Lost Her Case
A German AI image lawsuit involving an underwater dog photograph has become one of the clearest European examples of how courts separate a protected photograph from an unprotected visual idea. The Higher Regional Court of Düsseldorf rejected the photographer’s request to stop publication of a comic-style AI image derived from her photo.
The result sounds broader than it really is. The court did not declare that copyrighted photos may freely be uploaded to AI tools, that every AI transformation is lawful, or that AI-generated images automatically receive copyright. It decided a narrower question: did the published AI image recognizably reproduce the photographer’s own creative choices?
المصدر: Karl-Heinz Meurer / Wikimedia Commons, CC BY-SA 3.0, cropped for layout
Key takeaways
- The photographer lost because the AI image copied the motif, not the protected photographic expression.
- The protected elements included framing, perspective, lighting, focus and the realistic dynamic effect created by photographic technique.
- The AI image showed the same basic scene but changed the framing, body position, visual style and depth treatment.
- The court also found that the AI output had not been shown to be a new copyright-protected work because no meaningful human creative process was demonstrated.
- The ruling concerned publication of the AI output, not the legality of uploading the original image to the AI system or training a model on copyrighted photographs.
What happened in the German AI image lawsuit?
The applicant was a photographer who offered underwater photographs of dogs. One of her images showed a dog beneath the water surface reaching for a red toy. She had also edited the photograph after capture. The opposing party had previously cooperated with her, later uploaded the image file to an AI tool and caused the system to produce a comic-style variation.
The AI-generated image was then published on a website. The photographer discovered it on October 15, 2025 and, after an unsuccessful cease-and-desist demand, sought a preliminary injunction. She wanted the defendant prohibited from reproducing or making the AI image available to the public without her consent.
المصدر: Andrew Patrick Photo / pexels.com
This licensed stock photograph is illustrative and is not the image disputed in the case. The actual lawsuit concerned a tightly framed underwater dog portrait and a later comic-style AI variation.
The Düsseldorf Regional Court rejected the application on December 22, 2025. It regarded the AI result as a sufficiently independent free adaptation. The photographer appealed, arguing that the images were nearly identical, that the shared subject and motif should matter, and that courts should apply a stricter standard to AI-assisted plagiarism.
On April 2, 2026, the Higher Regional Court of Düsseldorf dismissed the appeal in case 20 W 2/26. The photographer also had to bear the costs of the appeal. The official decision lists a dispute value of €7,000.
Timeline of the case
| Date | Event | Legal significance |
|---|---|---|
| October 15, 2025 | The photographer became aware of the published AI-generated dog image. | This triggered the warning letter and urgent injunction proceedings. |
| December 22, 2025 | The Düsseldorf Regional Court rejected the requested preliminary injunction. | The first court considered the result sufficiently distant from the original. |
| March 3, 2026 | The appellate court addressed the apparent use of AI and the missing evidence about human creative decisions. | The defendant did not explain the prompts or other creative control in enough detail. |
| April 2, 2026 | The Higher Regional Court of Düsseldorf dismissed the appeal. | No infringement was found because the AI image did not reproduce the original’s protected creative elements. |
Why the dog photographer lost her case
1. Copyright protected the photograph’s expression, not the idea of a dog chasing a toy underwater
German copyright law protects personal intellectual creations. For a photographic work, that protection can cover creative decisions such as the chosen crop, camera angle, perspective, lighting, timing, depth of field and the balance between sharp and blurred areas.
The court drew a firm line between those choices and the underlying subject. A dog reaching for a red toy underwater is a motif. It is an idea or scene that other creators can also depict. Copyright does not normally give one photographer exclusive control over that subject in every visual style.
That distinction decided the case. The similarities identified by the photographer concerned the dog, the underwater setting and the red toy. The court regarded those as unprotected elements rather than as a reproduction of her individual photographic design.
2. The AI image changed the elements that made the original photograph distinctive
The original was described as a realistic and dynamic photograph in which the viewer mainly saw the dog’s head and the toy. The body receded because of the chosen perspective and blur. Its sense of motion depended on exposure, aperture and the photographic treatment of focus.
The challenged image looked different in the areas that mattered legally. It had a comic-like character, showed the dog’s full body and portrayed the animal reaching forward with both its mouth and exaggerated front paws. It lacked the original photograph’s realistic depth and dynamic photographic effect.
In other words, the AI image retained the narrative concept but not the concrete photographic decisions through which the photographer expressed that concept. The court therefore found no infringing reproduction.
المصدر: Terje Sollie / pexels.com
For photographic copyright, the legally important features are often the human choices visible in the finished image: composition, perspective, lighting, focus, timing and other concrete design decisions.
3. The AI output was not proven to be a new protected work—but that did not make it infringing
The appellate court disagreed with part of the lower court’s reasoning. Under Section 23 of the German Copyright Act, a “free adaptation” requires a newly created work. The defendant could not rely on that route unless the AI output itself reflected sufficient human creativity.
No one had documented the prompts, iterative instructions, settings or post-processing that produced the image. The court said AI-assisted output can in principle qualify for copyright protection when a human makes free and creative decisions that visibly shape the result. That influence can occur before generation, during an iterative process or through later manual editing.
Merely selecting one result from several AI suggestions is not enough by itself. Nor are numerous but open-ended prompts sufficient when the system still makes the actual visual decisions. Because the defendant did not explain any concrete creative choices, the court did not treat the AI result as his original protected work.
This did not hand victory to the photographer. A work can be unprotected itself without infringing another work. The decisive question remained whether the output copied recognizable protected elements from the original. The court said it did not.
4. The photographer’s application targeted the published output, not every potentially relevant act
The case was an urgent proceeding focused on stopping publication of the generated image. It did not decide every legal issue that could arise from the workflow. In particular, the court did not rule on whether uploading the original file to an AI service was an unauthorized reproduction under Section 16.
Legal analysis of the decision has emphasized that the upload itself could raise a separate reproduction question. But that act was not the basis of the requested injunction. Contractual restrictions from the former collaboration, unfair competition, data-protection questions or other claims were also not resolved by this copyright comparison.
What the ruling does—and does not—say
| Question | What the court decided | What remains open |
|---|---|---|
| Was the original photograph protected? | Yes. The court treated it as a protected photographic work. | The exact breadth of protection still depends on the creative elements of each photograph. |
| Did the AI output infringe that photograph? | No, because it reproduced only the unprotected motif and not the recognizable creative photographic choices. | A closer AI output that preserves framing, perspective, light or focus could produce a different result. |
| Was the AI output itself copyrighted? | Not on the evidence presented. The defendant did not show sufficient human creative control. | AI-assisted work may be protected when human creative decisions are documented and visible in the result. |
| Was uploading the source photo to the AI tool lawful? | The court did not decide that issue in this proceeding. | Uploading can raise separate reproduction, license and contract questions. |
| Did the court approve AI training on copyrighted images? | No. This was not a training-data case. | Dataset creation, text-and-data mining exceptions and model training are governed by other legal questions. |
Why the ruling matters for photographers and AI users
The German AI image lawsuit shows that “looks similar” is not a complete copyright test. Courts must identify which parts of an original express the author’s own creative decisions and then ask whether those specific elements remain recognizable in the challenged output.
For photographers, this makes evidence and claim selection critical. A claimant should be able to explain what is original about the image beyond its subject: the crop, angle, light, timing, focus, staging, color treatment and post-processing. Agreements with clients or collaborators should also address whether files may be uploaded to generative systems, used as references, transformed or republished.
For AI users, the decision is not a safe-harbor rule. Using another person’s image as an input can create separate legal risks even when the final output is visually different. A closer transformation may still copy protected elements, and contractual terms may prohibit uploading the source file. Businesses should keep records of licenses, source files, prompts, model settings, iterations and manual edits.
المصدر: Karola G / pexels.com
Written agreements can address AI uploads, reference-image use, derivative marketing assets, confidentiality and ownership before a creative collaboration ends in a dispute.
How the court approached copyright in AI-generated output
The judgment is notable because both sides effectively faced a human-creativity test. The photographer needed to identify the original photograph’s protected creative elements. The defendant, if he wanted to rely on the AI image as a new work and a free adaptation, needed to show that his own creative choices shaped the output.
The court described several ways human contribution might become legally relevant:
- Before generation: specific instructions or settings that make concrete design choices rather than leaving the result open.
- During generation: a documented iterative process in which the user deliberately steers composition and appearance.
- After generation: meaningful manual editing or post-production that gives the final image an individual human character.
The outcome must visibly embody those choices. Effort alone is not enough. A long prompt, many attempts or an expensive workflow does not automatically create copyright if the system remains responsible for the decisive expressive features.
This fits the broader European principle that a protected work must reflect free and creative choices of a human author. The Düsseldorf court applied the Court of Justice of the European Union’s December 4, 2025 guidance in the joined Mio and konektra cases when identifying protectable elements and testing whether they had been reproduced.
This was an output case, not an AI training ruling
Headlines about a “German AI copyright ruling” can blur several different legal stages. This dispute concerned a particular output created from a particular photograph and then published. It did not decide whether an AI company may scrape millions of images, whether a training copy falls under a text-and-data-mining exception, or whether an opt-out was effective.
Readers following the wider copyright debate can compare this decision with Zerlo’s overview of the
Could a photographer win a similar case?
Yes. The Düsseldorf ruling does not create a general immunity for AI transformations. A photographer could have a stronger claim where the generated image retains a distinctive crop, viewpoint, lighting pattern, depth-of-field treatment, staging choice or other recognizable expression from the original.
A different legal claim could also change the analysis. Unauthorized upload of the source file, breach of a license or cooperation agreement, removal or misuse of authorship information, unfair competition, trademark use or privacy rights may require separate evidence and remedies. The court even noted that allegedly adding the photographer’s identifier to the AI image concerned a different dispute that had not been properly pursued in the initial application.
FAQ
What was the German AI image lawsuit about?
It concerned an underwater dog photographer who sought to stop publication of a comic-style AI image created after her photograph was uploaded to an AI tool. The Higher Regional Court of Düsseldorf rejected the request on April 2, 2026.
Why did the dog photographer lose?
The court found that the AI image reused the unprotected motif of a dog reaching for a red toy underwater but did not reproduce the original photograph’s protected framing, perspective, focus, lighting and realistic dynamic effect.
Did the court say AI-generated images are automatically legal?
No. The decision was limited to the specific output and claims before the court. A different AI image may infringe when protected creative elements of the original remain recognizable.
Did the AI image receive copyright protection?
Not in this case. The defendant did not explain the prompts, settings, iterations or editing well enough to show that the output embodied his own free and creative human decisions.
Is uploading a copyrighted photo to an AI tool permitted in Germany?
The Düsseldorf court did not decide that question. Uploading may constitute a separate reproduction and may also breach a license, contract or platform rule. The answer depends on the authorization, purpose and applicable exceptions.
Does copyright protect a photographic motif?
Normally, copyright protects the concrete expression rather than the general idea or subject. A motif such as a dog chasing a toy underwater is not monopolized, but a photographer’s particular composition and technical-creative execution can be protected.
Is this ruling binding across the European Union?
No. It is a German appellate decision. However, it applied EU copyright principles and relied on recent Court of Justice guidance, so it may be influential in future European disputes.
Bottom line
The photographer lost the German AI image lawsuit because the court saw a shared idea but not a reproduction of her protected photographic expression. The AI image was also not proven to be a new protected work, yet that did not turn it into an infringement. The ruling therefore cuts in two directions: AI users cannot assume that prompting automatically creates copyright, while photographers must show that recognizable creative choices—not only the motif—were copied.
It is a significant decision, but a narrow one. It does not authorize AI training, settle the legality of uploading protected files, or immunize transformations that preserve an original work’s distinctive composition. Creators and businesses should document the entire workflow and assess each act separately.