(By You Yunting) On April 2nd, 2026, the Düsseldorf Higher Regional Court in Germany issued a final judgment regarding a dispute over an AI-generated image that mimicked a photographic work, holding that the defendant’s conduct did not constitute copyright infringement. Under China’s Copyright Law, such conduct would also fall outside the scope of copyright infringement. This judgment means that a significant portion of photographers’ business may be lawfully replaced by AI. The underlying cause is that the traditional idea-expression dichotomy in copyright law can no longer adapt to the creation logic of generative AI.
I. AI Repurposed the Idea, Not the Expression
Case summary: The plaintiff in this case is a photographer who once had a commercial cooperation relationship with the defendant, as a result of which the defendant had access to the plaintiff’s photographic works. Subsequently, the defendant uploaded the plaintiff’s photographs into an AI software called “X” and instructed the AI to generate an image similar in style and content to the plaintiff’s work. The plaintiff sued for copyright infringement, but the lawsuit was ultimately dismissed by the court in the final judgment.
The German court held: the plaintiff’s photographic work actually showed only the dog’s head and the toy, because through perspective and blurring, the dog’s body was blended entirely into the background, producing a realistic and dynamic image. And these very elements were missing from the accused image (which adopted a comic style). In the accused image, the dog’s entire body was clearly visible; it not only tried to hold the toy in its mouth but also attempted to grasp it with its forward-stretched front paws. This accused image lacked the dynamic effect created by the plaintiff’s exposure and aperture control. Based on this, the court found that the expressions in the two images were different. Therefore, the court ruled: The AI-generated image did not copy the elements of the plaintiff’s photograph under the protection of copyright law (such as shooting angle, lighting, and sharpness), but merely used the unprotected subject matter and idea (the scene of a dog chasing a toy underwater).
II. The Dilemma of the Idea-Expression Dichotomy in the AI Era
Actually, if this case occurred in China, Chinese courts would most likely also find the defendant not liable for infringement, because under the globally accepted idea-expression dichotomy, copyright law protects only the specific expression of ideas, not ideas, concepts, procedures, methods of operation, or mathematical concepts themselves. This principle aims to encourage the free dissemination of ideas and subsequent innovation, preventing anyone from monopolizing a broad field of creativity. China’s Copyright Law has also adopted this principle. In this case, the AI copied the idea of the original work. If ideas were copyrightable, the public’s creative freedom—for example, another person taking a photo of a dog underwater trying to retrieve a red toy ball—would be severely restricted.
In the era without AI, the idea-expression dichotomy was relatively balanced and effective. Because for a human to imitate an idea and create a new expression, considerable skill, time, and cost were required. For instance, for another photographer to imitate the “underwater dog” idea, he/she would also need to find a suitable location, train the dog, purchase underwater shooting equipment, and use his/her own photography skills to create the work. The final product would inevitably carry his/her own original expression. The market could accommodate two works based on the same idea but with different expressions, competing fairly.
However, generative AI has completely broken this balance. By training on massive datasets containing countless copyrighted photographic works, AI models have internalized human photography aesthetics, composition rules, lighting techniques, and even the personal styles of specific photographers. Users only need to describe an idea in natural language, and AI can instantly generate countless new image expressions that are technically and aesthetically sophisticated. In this process, AI has actually accomplished the precise extraction and commercial re-use of the most valuable part of the original work—the idea or concept, while cleverly avoiding mechanical reproduction of the specific expression.
III. The Commercial Foundation of the Photography Industry Is Being Hollowed Out
Although AI has improved the efficiency of image creation, it also hollows out the commercial foundation of the photography industry, an important content industry:
First, a substantial decline in commercial interests. AI has improved the efficiency of image creation exponentially with a dramatic cost reduction. Valuable ideas are efficiently exploited by AI at a low cost, while the uniqueness and commercial value of the expressions on which photographers rely for their livelihoods are diluted and overwhelmed by the massive amount of similar content generated by AI, thus severely impacting the market for photographers’ image sales. The legal protection of expressions appears powerless under the devastating impact of AI.
Second, a diminution in creative incentives. Under the traditional copyright protection system, photographers can expect reasonable economic returns from their creative inputs. That judgment means that AI can easily reproduce their works without constituting infringement, making it difficult for photographers to obtain corresponding returns from their creative efforts. Such free-riding may lead photographers to cut back on creative projects, or even leave the profession altogether and switch to other occupations.
A more serious problem is the systemic impact on the industrial ecosystem. Commercial losses suffered by photographers will lead to enrollment difficulties for photography educational institutions, business shrinking for image libraries and copyright agencies, and even market demand reduction for photography equipment manufacturers. Such systemic impacts may trigger unemployment waves among related practitioners and severely damage the entire photography industry chain.
Finally, the failure of the idea-expression dichotomy in the generative AI era is only a surface manifestation. There is a deeper issue that this case has not addressed: Since AI’s generative capabilities come from training on massive copyrighted works, should compensation be paid for training on photographers’ works? But currently, there is no clear answer. This is fundamentally a matter of legislating new rights to address the societal impact of AI technology. Before the development of AI technology is stabilized and relevant business models are established, legislative results are hardly expected within the next five years. Therefore, at this stage, the photographers and the industry can only look to self-help measures, such as shifting from selling copyright to selling experiences, providing more portrait services for individual users like children and women, and belatedly strengthening technical protection by reducing the resolution of photographic works published online or adding technical protection measures to prevent unauthorized AI training.
Lawyer Contacts
Disclaimer of Bridge IP Law Commentary
Short Link: