Midjourney and Its Future Implications for Intellectual Property regarding Content generated by Artificial Intelligence
The recent boom in generative Artificial Intelligence (AI) programs, which refers to applications that are designed to produce text, images, and computer programs, has brought quick and unprecedented development in the AI field. Amongst the many generative AI programs that have been launched in this past year, Midjourney, a generative AI service, which creates images based on natural language indications written by users, has stood out in terms of its exponential adoption and rapid growth (Midjourney, n.d.). With more than 15 million users and an average user base growth rate of 100,000 users a day, Midjourney has become one of the most important generative AI programs in the world (Dawood, 2023). This program is transforming the way we think about and know creativity and design invention, and consequently, it also has challenged the conventional application of Intellectual Property (IP) laws applied to the designs generated by the program.
The biggest question regarding how to rule AI-generated content, such as images generated by Midjourney, was whether they were protectable under current IP laws. At this stage, the answer by IP experts is that it depends on the local laws of the country, specifically regarding computer-generated content. For example, the United Kingdom (UK) law states that the author of a computer-generated work can seek limited copyright protection under the rationale that the author did necessary arrangements for the computer program to produce such work (de Reya, 2023). Hence, the UK considers that as long as there is some human input (program arrangements, such as prompts) in the AI-based work generation, authors may still seek limited copyright protection of the work generated by their arrangements.
United States (US) law initially took a different stance, where the law stated that only human-generated work parts in AI-generated work can be protected by copyright. An initial relevant question was how to distinguish with sufficient clarity between human-generated and AI-generated work. However, the US law resolves this clearly by defining that only works, where humans have arranged AI-generated works in a sufficiently creative way to consider the end work as an original work of authorship, can be protected by copyright. This specification further became clearer when the US law granted copyright protection to an 18-page graphic novel created mainly using Midjourney-generated images, which originally served as an example of work still generated by AI, but sufficiently arranged and modified to consider the end work original (de Reya, 2023).
However, in February 2023, the graphic novel got revoked its copyright protection for all the content and only got granted copyright protection for text, which was the only human-generated part of the novel. Hence, with this ruling, the US Copyright Office decided to generalize and state that Midjourney generative images cannot be copyrighted, no matter the arrangements done to them. Hence, the Office revisited its initial consideration that Midjourney works can be sufficiently modified or arranged to consider the end work an original one. This new ruling has significant implications for other generative AI programs, as experts consider that it is very likely that this type of ruling extends to other AI-generated works, such as text (e.g., ChatGPT, Bard), voice generation, etc.
This ruling has the potential to change the way creators or businesses use and navigate AI-generated works. Since they will likely not have copyright protection for their AI-generated works, they need to be more cautious when using this new technology. For creators, the challenge lies in finding other ways to monetize their AI-generated work. Some IP experts have suggested exploring IP protection through patents or trade secrets. Since patents technically protect the process through which works are created, experts think that there can be IP legal relief for creators through this IP area. However, it is still yet to be seen how local IP organizations and institutes will react to these potential new patents.
For companies using third-party agencies to create original works for them, this ruling implies that agreements of not using generative AI tools to create the work in question are vital if they plan to seek copyright protection for this. Since they may not be aware of their agency using generative AI programs, they might face surprises when looking for copyright protection if the local IP organizations found out that the work has AI-generated parts, and thus, the produced work by third parties is not eligible for copyright.
Another key element that the US Copyright Office clarified in their ruling was that the decisive consideration that inclined them towards this ruling was “the fact that Midjourney's specific output cannot be predicted by users makes Midjourney different for copyright purposes than other tools used by artists. (Brittian, 2023)” Hence, the Office provided certainty regarding the creations by any type of new tech tools that can be eligible for copyright, which are those where the artist has full control of the output generated by them.
For example, the well-known design hub called Canva, where users can access millions of templates from fonts, symbols, free-license images, and any graphic design, recently launched its AI studio, where platform users can speed up their graphic creation process through AI editing suggestions that range from changing colors of objects and symbols in the graphic to suggesting new images that they can input (Canva AI, n.d.). This process differs from Midjourney’s one in that it still gives full decision power to the user regarding which AI recommendations to include or not include. Thus, the user can predict throughout the creation process what will be the exact output of their AI-assisted creation. In this case, the US Office Copyright rationale of the ruling works perfectly to consider which creations using Canva cross the line of not being human generated even with the inputs human might give them, and thus, are not eligible for copyright protection, and which ones can still be considered human-generated ones.
Whether it is Canva or Midjourney, a certain issue is that IP laws remain yet to be tested on new disputes and filings concerning works created by generative AI programs. Both globally and locally, IP Offices will need to adapt and firmly decide which AI-generated content will be protected by specific IP laws. For now, the US has provided some clarity on the copyright realm concerning images and graphics generated by Midjourney given the graphic novel case. However, this ruling is only the beginning of many more that need and should come to give clarity to creators using these tools over the protection of their work.
It is expected that new rulings will come and clear off the ambiguous ground regarding other sectors of IP laws. For example, we are yet to see and learn more about how patent law can fairly treat the process of images and works in general created using generative AI tools, such as Midjourney.
References
https://www.mlyearning.org/midjourney-users-statistics/
https://www.midjourney.com/home/?callbackUrl=%2Fapp%2F
https://www.linkedin.com/pulse/new-ruling-holds-midjourney-ai-images-cant-what-does-j ackson-esq-/?utm_source=share&utm_medium=member_ios&utm_campaign=share_vi a
https://www.ibanet.org/AI-generated-works-an-IP-headache
https://www.reuters.com/legal/ai-created-images-lose-us-copyrights-test-new-technolog y-2023-02-22/
https://www.canva.com/features/ai-photo-editing/
By: Margarita Guerrero González