What Happened?

The claimant had created an image (see below) using a leading text-to-image AI model. A simple prompt had been used: “create a visual representation of two parties signing a business contract in a formal setting, for example in a commercial room or in a law firm office in Prague. Show only the hands.”

The generated image was sent by the claimant to the defendant, a Prague law firm, who then used the image on its website without obtaining consent from the claimant. The claimant then brought proceedings before the Court to: 

1. confirm the authorship of the claimant, and
2. prohibit the defendant from using the image. 

No Author to be Found

The Court’s key reason for rejecting the claim was that the claimant could not prove that he was the author of the image. The Court had required the claimant to provide evidence as to how the image was created, namely to prove who instructed the AI model and on the basis of which specific prompt (of a specific person) the AI model created the image. The Court concluded that the claimant had not managed to do so.

Neither of the parties disputed that the image was generated through AI. However, the Czech Copyright Act contains an express provision that an author may only be a natural person who created a work. The Court considered that the AI model itself cannot be the author and consequently this condition was not met.

The claimant had argued that the image was created by AI on the basis of his specific prompt, which justified the claimant’s copyright to the image. However, the Court found that the claimant had not actually proven that the image was generated on the basis of this prompt. Aside from his personal statement, the claimant had not produced further evidence to prove that the prompt led to the image. 

No Copyright as a Matter of Principle

Beyond the reasons leading to the dismissal of the claim, the Court also felt the need to comment on some aspects of copyright and AI in general. In the Court’s opinion, if the image was not created personally by the claimant, but generated by AI, it cannot as a matter of principle be protected by copyright. The Court’s reasoning was that such an image is not the result of creative activity of a natural person – an author. Consequently, the Court concluded that the image is not a copyright-protected work, let alone a work created by the claimant.

The Court also considered that the prompt itself could only be regarded as a theme or idea for a work, neither of which can be protected by copyright. The Czech Copyright Act specifically lists themes, ideas and similar more abstract concepts as excluded from copyright protection, which is a principle generally recognised internationally. 

How Does This Link to Global Developments?

The Czech judgment ties into the international discussion on copyright protectability of AI-generated output. The approach taken by the Court in requiring human creative activity is comparable to the position taken so far by courts in the United States. Most notably, the US District Court for the District of Columbia in the Thaler case in August 2023 (see judgment here) denied copyright protection of AI-generated output, arguing that human authorship is a fundamental requirement.

The US Copyright Office has adopted a similarly cautious approach in its March 2023 policy guidance on works containing AI-generated materials (see here). It emphasises that users do not exercise ultimate creative control over how AI systems interpret prompts and generate material. 

A more AI-friendly approach has so far been seen in China. In Li vs Liu, in November 2023, the Beijing Internet Court ruled in favour of copyright protection of an image generated by the claimant using the Stable Diffusion Gen-AI model. The claimant had used about 150 prompts and negative prompts and set various parameters before the final image was generated. The court held that the image is a direct result of the claimant’s intelligent input and individual expression, and therefore copyright eligible (see our report here). An AI-generated image is considered copyrightable if it shows original intellectual input from a human. (See our earlier article on this decision in China: Copyright Protection for AI generated works – Recent Developments). 

Looking Ahead

The approach taken by the Court illustrates a further hurdle in the debate on the possibility of copyright protection of AI-generated works and its enforcement. Even if the complexity of a prompt could (theoretically) justify copyright to the output, AI users must also be prepared to evidence the causality of how their prompt (i.e., their creative choices) led to the creation of the output in question.

The Czech Court’s approach is consistent with a strict requirement on copyright through creative choices by a human author, which is likely to be expected from courts elsewhere in the EU as well. However, the Czech case concerned only a very simple prompt and further EU case law has not yet emerged. It therefore remains to be seen whether this strict approach could be overcome in cases of detailed prompts and overwhelming human input, coupled with AI-assisted execution. 

The full text of the Czech decision can be found here (in Czech only).

If you need more information or further guidance in this area, please contact Vojtěch Chloupek and Martin Taimr. 

WRITTEN BY

Vojtech Chloupek

PARTNER

Martin Taimr

ASSOCIATE