First EU court to consider who holds copyright in AI generated works suggests room for ownership if prompts are sufficiently creative in parallels with a recent Beijing Internet Court decision.
Blog Intellectual Property Notes

Can an AI-generated works be protected by copyright? The first answer from an EU court comes from a Czech decision which ultimately found that AI could not be an author of copyright but seemed to imply that, as has recently been held in a case before the Beijing Internet Court, a sophisticated prompt might be sufficient to allow the human creating the prompt to have copyright in the prompted AI's output.

The case

The claimant created an image using DALL-E, an AI system developed by OpenAI that can create images from text instructions, so-called "prompts", given by the user. The prompt provided to the AI was fairly simple in this case: “create a visual representation of two parties signing a business contract in a formal setting, such as a conference room or a law firm office in Prague. Show only the hands.”

Being satisfied with the image created by DALL-E as a result of this prompt, the claimant then published this AI-generated picture on his website, making it freely available. The claimant later became aware that the defendant had copied the picture and published it without any authorisation on its own website.

The claimant filed a lawsuit for copyright infringement before the Municipal Court of Prague seeking an injunction for the removal of the image from defendant's website, on the grounds that the claimant was the author of the image and owner of the copyright therein and that the reproduction by the defendant was unauthorised and constituted copyright infringement.

The decision

The Czech Court dismissed the case on procedural grounds finding that the claimant failed to meet the required burden of proving that the image was in fact the claimant's own creation. In the opinion of the Court, the claimant did not present sufficient evidence to show that the image was generated on the basis of the claimant's prompt. Therefore, the Court could not assess who the author of the image was and, consequently, held that the claimant did not have standing to claim copyright infringement as this could only be claimed by the author of the creative work.

More significantly, the Court stated that a work of authorship must be the unique result of the creative activity of a natural person, which could reasonably entail that:

(a) unless the creator can demonstrate that the generated image is the result of their unique creative contribution, the image cannot be held a work of authorship; and

(b) the AI, not being human, cannot be the author of a work under national copyright law.

Following this reasoning, it seems that the definition of the prompt could amount to a "creative activity" and that, therefore, if a person could prove to a court that the AI generated work is the result of a sophisticated prompt, the AI generated image could be granted copyright protection.

A comparative view

Only months away from the final approval of the EU AI Act, the Czech case is the first national decision issued in a EU Member State on the issue of copyright protection of AI-generated works. This judgment follows two prior decisions issued last year, two from the United States Copyright Office (USCO) (with one decision appealed and subsequently upheld by the District Court of Columbia) and one by the Beijing Internet Court in China, both of which addressed the same question but with different outcomes:

• US: The USCO held that an AI-generated work is not eligible for copyright protection. The Office often stated that "Human authorship is a bedrock requirement of copyright” because "the Office will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author." (U.S. Copyright Office, Compendium of U.S. Copyright Office Practices sec. 313.2 (3d ed. 2021)). These principles found application in two cases decided in the past years by the Office.
  • Zarya of the Dawn: The USCO rejected the application of the copyright registration application presented by Ms. Kristina Kashtanova for her comic book Zarya of the Dawn, the graphics of which were realised using AI-generated images. The USCO held that, while "the selection and arrangement of the images and text in the comic book were protectable as a compilation," the images generated by the AI system were not original works of authorship protected by copyright because "users lack sufficient control over generated images to be treated as the “master mind” behind them," regardless of the efforts the user put into researching the perfect prompt to obtain a rendition as close to the user vision as possible.
  • A Recent Entrance to Paradise: In a similar case, the USCO rejected the application made by Dr Thaler, to have copyright registered for an image made by his "Creativity Machine" AI image generator with the AI as author. On appeal, the District Court of Columbia upheld the USCO's finding, holding that Dr Thaler (whose was the AI said to have generated the images for which copyright registration was sought) had “…correctly observe[d] that throughout its long history, copyright has proven malleable enough to cover works created with or involving technologies developed long after traditional media of writings memorialized on paper … Copyright has never stretched so far, however, as to protect works generated by new forms of technology operating absent any guiding human hand, as plaintiff urges here. Human authorship is a bedrock requirement of copyright.”, (See our blog post on the USCO's decision and its appeal here).
• China: The Beijing Internet Court, faced with a similar set of facts, reached the opposite conclusion, holding that an AI-generated image can indeed be protected by copyright law. It held that this was because the outcome of the algorithm can be influenced by the way a prompt is drafted and refined as well as by how the instruction parameters of the AI system are set, therefore resulting in the unique expression of the human author.

In the Chinese case, the issues were addressed by way of litigation. Mr. Li had used an AI system to create an image that he then posted on his personal profile on Little Red Book, a popular social media platform. Upon discovering that Mr. Liu had used the image without permission in a publicly available article, Mr. Li sued for copyright infringement before the Beijing Internet Court and sought the removal of the image and compensation. While, the Court of Beijing found that the image was the result of a trial-and-error process aimed at achieving a reliable rendition of the user's vision; unlike the USCO, the Chinese Court held that the selection and arrangement of the different prompts, relevant parameters, reflected the claimant’s aesthetic choice and personal judgment and that, therefore, the use of AI is comparable to the use of any other tool. The Court concluded that since the final image is thus a reflection of the original intellectual investment of the human mind, an AI-generated image could be protected by copyright.

These decisions demonstrate the struggle that legal systems are having with respect to copyright protection of AI-generated work. The USCO and court have taken a restrictive approach of not allowing copyright protection for AI-generated work, while a Chinese court has used a broader approach to conclude that AI-generated work can be protected under copyright as a creative expression of the user.

Now enter the EU as a mid-point. Unfortunately, the case was such that the Municipal Court of Prague could not make any substantial conclusions and had to dismiss the case on procedural grounds. However, on one reading, the Czech Court seemed to be leaning towards the US's restrictive approach, stating that the image in question was created by AI and could not by definition be a copyright work without a human author. Nonetheless, the Court of Prague underlined that the specific prompt on the basis on which the image was created was not fully established. This suggests that the Court recognises the relevance in principle of a prompt, perhaps even to the extent of considering the prompt as a unique creative contribution, an intellectual investment, a reflection of the user's personality judgment, as was applied by the Chinese court.

Conclusions

While it is clear that the AI, being not human, cannot itself be the legal author of a work, there is still much debate about the possibility of the subsistence of copyright in (or "copyrighting") an AI-generated work. If the AI, like a camera, can be considered simply a tool and the AI-generated art a simple work of art, then the process of copyright subsistence and ownership could be much simpler. An alternative view is that AI is something so revolutionary that it requires a whole new legal framework. The US and Chinese decisions are not necessarily conflicting but certainly provide contrasting outcomes for now. It is disappointing that this first EU case was not able to come to any definitive conclusions on the question. What would the Prague court have decided if the claimant had met the burden of proof? However, the investment in designing a unique prompt may turn out to be the inspiration for AI users to create a path to copyright protection of the AI's outputs.

For more on AI and copyright see our series The IP in AI and our additional reports here.