Introduction
 
ChatGPT was just the beginning. Since 2022, generative artificial intelligence (AI) tools have infiltrated every aspect of our lives, redefining how we create, share and control content. With the rise of multimodal models¹, content creation has become richer and more interactive, with some offering built-in features to suggest refinements or alternatives to user-inputted prompts. These advancements have fundamentally altered the art and science of prompting – a seemingly simple yet profoundly impactful process central to our interaction with generative AI.
 
But alongside these innovations, new considerations for intellectual property (IP) protection have emerged and one area that has garnered attention is the strategic use of prompts and whether we can use them to manage IP infringement risks and assert ownership of the resulting outputs. This article explores these issues, identifying key strategies for navigating the complex intersection of AI and IP.
 
The Power of Prompts? Case Studies from the US and China
 
Recent cases in the US and China highlight the significance of the human element in AI-assisted creations, rather than just the final output. While they do not provide a blanket endorsement of prompts as a basis for claiming IP rights over AI-generated content, the courts did scrutinise the prompts amongst other aspects to determine the human expression throughout the creative process.
 
Zarya of the Dawn ²
 
In 2023, the U.S. Copyright Office garnered significant attention with its decision concerning Kris Kashtanova’s use of Midjourney to generate images for a graphic novel called ‘Zarya of the Dawn’. Initially, the novel was granted copyright registration. However, upon learning that it included AI-generated images, the Copyright Office cancelled the original registration but allowed a new registration more narrowly focused on the contributions of the human author, namely the text of the graphic novel and the selection of arrangements of the AI-generated images.³  In its examination, the Copyright Office determined that Kashtanova had "influence[d]" the generated image through prompts,  but there was still too much “distance” between Kashtanova’s input and the output for the artist to have enough control over the generated images to be considered their “mastermind.”⁴ The images therefore did not qualify for copyright protection.
 
Li v Liu
 
In comparison, the Beijing Internet Court took a different stance in Li v. Liu.⁵  Mr. Li sued Ms. Liu for infringing his rights of authorship and dissemination. He had inputted detailed prompts into Stable Diffusion, a generative AI system, to generate a picture of a young woman and published it on the Xiaohongshu social media platform. Ms. Liu, a blogger, used a copy of the image in an article published on Baijiahao, another online platform, without permission, attribution or compensation to Mr. Li. Ms. Liu also removed Mr. Li’s user ID and watermark. Mr. Li sued, seeking a public apology, removal of the image from Ms. Liu's online site, and monetary compensation.
 
Under Chinese copyright law, a work must be an intellectual creation (by a human) to be protected by copyright. On this criterion, the court held that Mr. Li provided substantial intellectual inputs throughout the picture-generation process, including: 

As a result, the court determined that the subject picture had reflected Mr. Li’s intellectual input, thus meeting the criteria of “intellectual creations”.
 
Both cases illustrate that whether the use of AI can satisfy the requirements of copyright law will heavily depend on the facts and whether the prompter can prove that the resultant output is not just a mechanical intellectual achievement but one that possesses originality of the author.⁸ Building on this, a thoughtful prompt strategy can be a valuable component of a comprehensive IP management approach for organisations leveraging AI content generation.
 
Enhancing IP Rights Through Prompting: A Simplified Guide
 

 
Conclusion
 
A company’s IP strategy for AI-generated content must carefully consider other aspects such as the contractual terms to determine IP ownership (such as whether the human prompter retain rights to the IP under the terms) and content protection measures such as adding disclaimers or watermarking AI-generated content to reinforce ownership and deter unauthorised use. However, when done thoughtfully, prompting or prompt engineering can become a key factor in establishing ‘originality’ in AI-generated works. In that sense, the humble prompt, once a mere input, is potentially a powerful tool in IP strategy. But it is also important to recognise its limitations. Courts are likely to examine the totality of the human’s creative input and decision-making process rather than just the prompts used.
 
Article by Natalie Lim (Partner) of the Technology, Media, and Telecommunications Practice of Skrine. 
 

---