The 'Prompter' as Copyright Author? Lessons from Munich

For the first time, a German court has addressed the "Prompter's" claim to authorship in the output. The District Court of Munich (AG München, Judgment of Feb 13, 2026 – 142 C 9786/25) offers a foundational blueprint for determining when the human-machine prompt crosses the threshold of copyright protection when it comes to the output.

The Case

The plaintiff created three logos (a handshake/bell, an envelope, and a laptop) using a text-to-image generative AI system.   

Image

He argued that his extensive, iterative prompting – including a 1,700-character instruction for a single logo – constituted a "personal intellectual creation" (§ 2 Abs. 2 UrhG). He sued for infringement when a third party utilized the logos without his permission.

The court dismissed the claim, denying copyright protection to all three images. The key legal takeaways from the judgment are the following: copyright protection depends on whether the product "reflects the personality of its author by expressing their free creative decisions." (para. 18). In a phrasing unique to the current landscape of AI jurisprudence, the court stated that for copyright to subsist, the AI model must be "closer to a mere tool (Hilfsmittel) than to an independent instrument of creation." (para. 21). Also, it is not enough to merely "trigger" a process or select from multiple outputs. In Levola Hengelo (C-310/17), the cheese taste case, the CJEU ruled that the work itself must be identifiable with "sufficient objectivity." Here, the Munich court takes that European requirement and applies it upstream to the prompt. The court’s logic is: If the prompt is so vague ("make it artistic") that the AI has a billion ways to interpret it, the human has not "objectively defined" the output. The output is then a product of the machine's randomness, not the human's identifiable intent. Protection only arises if the creative elements within the prompt "dominate the output so much that the object can be seen as the author's own original creation." (para. 19-21). The court held that "merely generally formulated, open-ended instructions" leave the actual "design decision" to the AI, thus failing the test for authorship.

"Nice! Can you make the whiteskin hand more feminine?" – Why General Prompting Fails

The court analyzed the plaintiff’s iterative "refining" and found it wanting. Even when the user gave specific follow-up commands, they failed for two reasons: Phrases like "make the bell look more artistic" or "add a more realistic touch" were deemed "ergebnisoffene Anweisungen" (open-ended instructions). These are mere "briefs" to a machine that then makes the actual design decisions. The court noted that correcting obvious AI errors or requesting minor adjustments ("make the hands a bit more filigree") are routine technical adjustments rather than creative leaps. As the court stated in para. 27: "The subsequent mere selection of one product from four proposals is not sufficient."

Analysis: Why the Decision is Correct

The Munich court’s decision is fundamentally sound. To understand why, we must look at the two systemic hurdles of AI authorship.

1. The “If” Question: The Anthropocentric Wall

Copyright law is anthropocentric; it places the natural person at the core. This is evidenced by the post-mortem auctoris protection and moral rights. Machines or corporations cannot be authors. While the European Parliament's 2017 Resolution on Robotics famously floated the concept of 'electronic personhood', this idea ultimately failed to gain traction in copyright law. This is because authorship is inextricably linked to human intellectual intent. Unlike a natural person, AI lacks the autonomous creative agency required to 'express their personality' through free and creative choices, a core requirement of the CJEU’s Painer (C-145/10) and Cofemel (C-683/17) jurisprudence.   

Jurisprudence at the EU level – specifically Cofemel (C-683/17) and Brompton Bicycle (C-833/18) – reinforces that a work must "reflect the author's personality." As AI cannot be attributed a personality, it creates a systemic challenge for the direct protection of autonomous AI outputs.

2. The “How Much” Question: Randomness and Control

AI may serve as a sophisticated digital paintbrush; much like a photographer uses a camera to capture a specific vision, the artist uses the AI as a tool to manifest a pre-defined, creative intent. The decisive question is: how much of the human is in the output? Sufficient influence is established when the human concretely determines the essential features, and the AI serves merely as a tool to execute that design.

In this regard, we must differentiate between unprotectable style/ideas and protected concrete design. Vague ideas are public domain (BGH GRUR 1995, 47 - Rosaroter Elefant), and an artist's style is not protectable (BGH GRUR 1995, 668 (671) - Emil Nolde). For input to lead to a protectable output, it must outline the form and character of the final work, exerting a "concrete-formative influence" (the term used in copyright literature). Interestingly, the presence of randomness does not automatically disqualify a work.

To see where the Munich logos fell short, we should look at a world-famous example of "calculated" art: the south transept window of the Cologne Cathedral.   

Image

See image (photo by: Matthias Jung (KNA)

It was designed by Gerhard Richter, who is widely considered one of the most influential living painters in the world. In 2007, the Cathedral – a UNESCO World Heritage site – asked Richter to design a massive new stained-glass window to replace one destroyed in WWII. Instead of painting a traditional scene with figures, Richter decided to create a "pixelated" sea of color. To do this, he did not pick every single spot by hand; he used a computer program to help him. Even though a computer calculated the arrangement of the 11,263 glass squares, the window is a protected work of art. Why? Because Richter remained the "boss" of the process: he personally selected 72 specific tones to match the 700-year-old medieval windows already in the church. He told the computer exactly how to mirror the patterns and where the squares should go. When the computer's "random" result accidentally created shapes that looked like numbers or faces, Richter stepped in and manually changed them to fit his vision. The "spirit" of the window is Richter's because he controlled the framework and made the final aesthetic decisions. the Higher Regional Court of Cologne (OLG Köln, 2025) recently demonstrated a more settled approach. In the 'Richter Window' case, the court did not even 'dive' into the work criteria; it simply confirmed the lower court’s finding that the work – despite being computer-aided—was 'undisputedly a protectable work' under Sec. 2(1) No. 4 UrhG. This suggests that where 'framework control' is documented, German courts now treat the computer as a mere tool, bypassing the 'Anthropocentric Wall' entirely.

3. Why the Munich Logos are not "Richter Windows"

The Munich plaintiff's intervention was fundamentally different. Unlike Richter, who set a complex framework and aesthetic "bounds" for the machine, the plaintiff provided "general instructions" and then chose from the results. As the Munich court correctly noted, subsequent concrete instructions are not enough to trigger a copyright claim in the output if they do not go beyond general ideas. Correcting a skin tone or asking for a "feminine" hand is a technical correction of a machine's proposal, not a "concrete-formative influence" on the overall expression.

4. Contextualizing "Generic" AI: The Path to Protection

However, we must differentiate when it comes to post-processing. While mere selection from a large number of outputs does not establish protection, the creative organization and contextualization of AI content can. In the US Copyright Office case of the comic Zarya of the Dawn, the individual AI images remained unprotected, but the author’s creative organization into a sequence with original text qualified as a protectable human contribution. Consider also      what I would call the “Unitary Work of Art principle” ("Gesamtkunstwerk-Prinzip”): Transforming unprotectable AI-generated raw material into a complex, thematic physical installation effectively 'alienates' the generic imagery. By integrating these outputs into a curated, multi-sensory environment, the artist provides the necessary creative contribution to form a new, protected artistic work. In this context, the individual AI image is merely a component of a larger, human-directed intellectual creation.      

Final Perspective: The Limits of the Munich Logos

In the Munich case, the plaintiff’s logos failed because they never left the realm of general instructions. Even a 1,700-character prompt can remain a "briefing" rather than a "creation" if it merely lists requirements (e.g., "modern style", "blue colors") without determining the actual formative expression.

For future cases, we must remain technology-open: if a user "forces" a specific, identifiable form through highly complex, non-random prompt chains that pre-empt the AI's choices, the AI may finally become the "sophisticated paintbrush" the law requires.

Photo by Samuel Hagger on Unsplash