One of the interesting questions related to generative AI has been the copyright protection granted to content created with it, such as logos.
This topical question was addressed in European case law last year by a court in Prague. The case concerned an image created by the claimant using the DALL-E AI tool, the copyright protection of that image, and the potential liability of the defendant law firm for alleged infringement. In the case, the court placed the burden of proof on the claimant regarding the details of how the image had been created. As the claimant was unable to provide evidence beyond his own account, the court ultimately found that the claimant had failed to prove that he was the author of the image, and that the AI-generated image was not a work resulting from the creative activity of a natural person in a way that could qualify for copyright protection.
A similar line has now also been followed in a decision by the Local Court of Munich, which concerned the copyright protection of three AI-generated images.
What was the case about?
In short, the claimant had used a generative AI tool to create three different logos and published them on his own website:
The text prompts used by the claimant to generate the images were, on the one hand, fairly straightforward, but at times also quite detailed. The longest prompts exceeded 1,700 characters.
The defendant had then copied the logos and published them on its own website. In terms of copyright law, the defendant was therefore alleged to have reproduced the works and made them available to the public without the claimant’s consent.
In the dispute before the court, the claimant argued that the images he had created were protected under German copyright law and that the defendant had infringed his rights. The claimant compared the role of the AI user and the AI tool to the work of a sculptor, where the sculptor, meaning the AI user, gradually moves towards the final work using a chisel, meaning the AI tool. The defendant’s view, in short, was that the human only acts as a generator of ideas or as a client, while the generative AI acts as the actual creator.
How did the court decide the case?
The court stated unequivocally that the logos produced by the claimant using AI were not protected under copyright law.
In its reasoning, the court stated that the concept of a “work” under German copyright law is, in accordance with the established case law of the Court of Justice of the European Union, an autonomous concept of EU law that must be interpreted and applied uniformly in all Member States. The concept consists of two elements. First, the subject matter must be original in the sense that it is the result of the author’s own intellectual creation. Second, only an expression of such creative intellectual effort can qualify as a work.
For the subject matter to be considered original, it must reflect the author’s personality by expressing the author’s free and creative choices. Conversely, the subject matter cannot be considered sufficiently original if its creation is dictated by technical considerations, rules or constraints that leave no room for creative freedom.
According to the Munich court’s decision, creating a copyright-protected work with the help of AI requires that the role of AI in the creative process is closer to that of an assisting tool than an independent creative instrument. In other words, the AI must function more as a tool or as part of assisted use. By contrast, the court did not consider the length of the text prompt or the fact that the AI tool was used under a paid subscription to be relevant in its assessment.
What can be concluded from the decision?
The decision reinforces the general principle that AI-generated content, such as a logo, does not as a rule enjoy copyright protection. Copyright protection can only be granted to the result of the creative intellectual effort of a natural person and to an original work, as confirmed in established EU-wide case law.
In the assessment, decisive weight is not given to the time or effort spent on the work, such as the length of the text prompts or whether the application used was a paid version. This applies, for example, to the claimant’s 1,700-character prompt or the premium version of the AI application.
In essential terms, the decision concerns the relationship between a human author, their creative activity, and the use of generative AI. Although the outcome of the decision is not surprising, it may provide interpretative guidance on the requirements for obtaining protection. However, it is possible that courts will in the future assess situations where generative AI is more clearly used as a tool rather than merely as an automated generator. The interesting question will then be where the boundary lies: when does AI function as a tool, similar to Photoshop tools, rather than as the creator, and in what situations does AI-assisted creation meet the threshold for copyright protection?
At present, the main issue in the creative use of AI appears to be that there is not a sufficiently objective and unambiguous causal link between text prompts and AI-generated outputs. In other words, we do not fully know what kind of result a text prompt will produce in each situation.
In the meantime, it is worth remembering that logos created with AI can also be registered as trademarks.