Munich District Court AI-Generated Logo Ruling / Thoughts on the Boundary Between Prompt Input and Copyrightability

I. Judicial Judgment Regarding the Copyrightability of Generative AI Outputs

The question of what requirements must be met for images created using generative AI to be protected under copyright law is one of the major challenges facing current legal systems. In this regard, the Munich District Court in Germany issued a ruling on February 13, 2026, denying the copyrightability of a logo created using generative AI [Note 1]. This judgment establishes a clear standard that even if a user provides detailed instructions to an AI or makes repetitive corrections, it cannot be recognized as a copyrighted work as long as it does not go beyond the scope of commissioning work to an external party.

1. Background and Issues of the Case

The plaintiff created three logos using generative AI (referred to as "C" in the judgment). The designs were a handshake between people of different skin tones with a ringing bell, an envelope depicted in front of a building with columns, and a laptop with a book featuring a paragraph symbol floating in front of the screen [Note 1]. The plaintiff used these on their own website, but after the defendant, an acquaintance, posted the logos on their own company's website without permission, the plaintiff sought an injunction and removal on July 3, 2025, and filed a lawsuit after negotiations failed [Note 1].

In court, the plaintiff argued that they had embodied their own spiritual creation through detailed instructions and repetitive corrections, and that the AI was merely a sophisticated tool. In particular, they compared the repetitive corrections to the work of a sculptor carving a statue from stone step by step [Note 1]. In response, the defendant argued that the selection, combination, and design of visual elements were all performed automatically by the AI, and that the user's contribution was merely to trigger the technical process through text input. Furthermore, the defendant pointed out that the very nature of generative AI, which allows anyone to generate a large number of complex works with minimal effort, indicates that creative contributions are attributable to automated generation [Note 1].

2. Criteria for Determining Copyrightability

The court first stated that the concept of a copyrighted work is an autonomous concept of EU law, and according to established ECJ case law, originality is recognized only when it reflects the author's free creative choices. If creation is determined by technical considerations, rules, or other constraints, copyrightability is denied [Note 1].

Furthermore, the court ruled that whether an AI-generated work has copyrightability depends on the extent to which human creative influence is exerted, despite the software-controlled process flow. If a human intervenes subsequently and sequentially during the prompting process, and the prompt creator's personality is reflected in the output, copyright protection is conceptually possible. However, it is necessary for human influence to shape the output in a "sufficiently objective and clearly identifiable form," and the requirement is only met when the creative elements incorporated into the prompting dominate the output to the extent that the entire object can be regarded as the author's own original creation [Note 1].

The point made by Leistner (GRUR 2025, 1123, 1132), cited by the court, is also direct. The formulation that copyright protection applies "when the use of an AI model is closer to the character of a tool than as an independent means of creation" [Note 1] clearly shows the conclusion of the tool theory. The observation by Dreier/Schulze/Raue [Note 1] that copyright does not protect investment, time, or diligence, but only the results of creative activity, also forms the basis of this judgment.

II. Quantity and Qualitative Evaluation of Prompts

1. Divergence Between Prompt Length and Creativity

Regarding the envelope logo, the plaintiff had "created and tested" a complex prompt spanning 1,700 characters. However, the court dismissed the claim, stating that the magnitude of that effort did not constitute the result of creative activity under copyright law. In terms of content, descriptions such as "design an original and abstract logo," "modern, minimalist, and clearly original design," and "clean flat design" were deemed generic and did not indicate decisive creative judgment regarding the appearance of the specific output, concluding that the final design judgment was delegated to the AI [Note 1].

I believe this interpretation is appropriate. No matter how long the instructions are, as long as they are functionally equivalent to a work order given to a human designer, they cannot be evaluated as the creation of expression. The court's assessment that "a prompt is no different from a written purchase order" [Note 1] strictly maintains the question of "who formed the expression" in determining copyright ownership.

2. Legal Nature of Repetitive Corrections

In the process of creating the handshake and bell logo, the plaintiff repeatedly gave additional correction instructions to the output presented by the AI, such as "I want the fingers to be white-skinned" and "I want the bell to be more artistic." However, the court judged that these correction instructions were merely error corrections or fine-tuning, and that they did not reach the level of intervention that dominates the expression, as the decisive parts of the design were left to the AI's algorithm [Note 1].

The instruction "I want the bell to be more artistic" itself clearly indicates the abandonment of the core choice of expression. Even if a human specifies a particular style or element, if the AI's automated generation process is responsible for the formation of the final expression, the conclusion that human involvement is merely triggering a technical process is, in my opinion, unavoidable. Regarding the laptop logo, the plaintiff's instructions were limited to about two lines of description, and it was found that there were no decisions that creatively shaped the AI's output [Note 1].

III. Some Considerations

The standard clarified by this judgment has significant practical implications in that the quantity, precision, or repetitiveness of prompts is not the deciding factor for copyrightability. The plaintiff created a detailed 1,700-character prompt and repeatedly made corrections. Even so, copyrightability was denied based on the criterion of whether the prompt contained creative decisions sufficient to "dominate" the output, rather than the magnitude of the effort.

However, this judgment does not completely deny copyrightability for AI-generated works in general. The court explicitly acknowledges the possibility that copyright protection could be established if the creative elements of prompting dominate the output. The formation of specific criteria for what kind of involvement can be affirmed as "domination" remains a task for the future.

In discussions under Japanese copyright law as well, the question of how to interpret the requirement of "creative expression of thoughts or emotions" (Article 2, Paragraph 1, Item 1 of the Copyright Act) regarding the copyrightability of AI-generated works is at stake. I believe the standard of "creative decisions sufficient to dominate the output" in this judgment can serve as a strong reference point in comparative law.

IV. Conclusion

As the use of generative AI becomes commonplace, discussions surrounding the copyrightability of AI-generated works are beginning to move forward in various countries. Although this is merely a ruling by a single district court in Germany, the logic of this judgment, which carefully applied the interpretive framework of EU copyright law, is of reference value in predicting the future direction of legislation and the judiciary. The conclusion of this ruling—that as long as the details of the expression are left to the AI, it is difficult to acquire the status of an author in the legal sense, no matter how much effort is put into it—is a legal reality that everyone who uses generative AI in their business or creative activities should be aware of. For details, please refer to the original text.

(Magazine) "AI and Law - Random Thoughts"

* Please refer to the following for the table of contents

note General Terms of Service Article 3, Paragraph 2, First Sentence
3.2 The copyright of digital content created by creators belongs to the creator.