Picture this scenario: An experienced artist, known for his creative fireworks displays, is preparing a grand show at a stadium. To achieve an elevated level of artistry, he collaborates with his AI assistant — a text-to-image model that generates digital images from natural language descriptions known as “prompts.”
The artist trains this AI assistant using his earlier work to ensure that it understands his style. The AI assistant also possesses knowledge the artist doesn’t have — in fields like philosophy and psychology — which it incorporates into its designs to suggest new visuals. The artist reviews the AI assistant’s proposed images, selects one, makes some adjustments and works with the technical team to ensure a smooth launch of the fireworks.
![Jiaying Jiang](data:image/svg+xml,%3Csvg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 620 620"/%3E "Jiaying Jiang")
But then, legal questions arise. Are the images provided by the AI assistant copyrightable, especially when an artist trains AI using their own work? Who owns what rights? What about the prompts and digital files for the final fireworks display? The current copyright law has a hardline rule about AI, which leaves work contributed by AI unprotected. However, as artists continue using AI in their work, this creative collaboration needs to be protected.
So, what are the current copyright rules that artists have to follow? First, human authorship is a prerequisite for copyright protection, which means that animals and machines cannot be authors. Second, to be protected under copyright law, a work must be original, meaning the author must create it independently and it must exhibit some creativity. Third, copyright protection does not extend to ideas or functional elements in a work but only to the creator’s unique expression of those ideas. Fourth, the fair use of a copyrighted work does not constitute copyright infringement.
This existing copyright law is ill-suited for GenAI, unfortunately. Even if GenAI greatly contributes to the creative process, it will not be considered an author, and the requirement of originality is hard to meet. Artists provide initial prompts, parameters or specific instructions, and AI generates the final output, complicating the determination of whether the work is independently created by AI or sufficiently shaped by human creativity. The rule of protecting expression may not fit because the expression is contributed by AI through generating images and words, while artists only provide ideas and high-level concepts; this results in neither the artist nor the AI assistant being protected. GenAI often requires large datasets that may include entire copyrighted works, raising questions about whether the use can be justified under fair use doctrine.
Spend your days with Hayes
Subscribe to our free Stephinitely newsletter
Columnist Stephanie Hayes will share thoughts, feelings and funny business with you every Monday.
You’re all signed up!
Want more of our free, weekly newsletters in your inbox? Let’s get started.
Although the U.S. Copyright Office has hinted that a user owns the copyright (at least in theory, if they give sufficiently detailed instructions to the AI assistant that helps shape the ultimate work), in reality, the Copyright Office has never identified a user who has met this requirement to grant copyright. If we decide, as the Copyright Office has, not to protect AI-generated work at all, copyright law itself will slide into irrelevance as more and more creativity is bound up with AI.
Legal scholars have proposed a few solutions. Some argue that human artists should own the copyright of AI-generated work since AI systems do nothing creative and are tools like cameras and Photoshop. Other scholars propose that the company that owns or deploys the AI should be recognized as the author and owner of whatever it creates, and some scholars suggest that AI itself should be considered the author. There are also scholars who think that AI-generated creative work may create a new ownership right, or that no one should own the rights and that the works should fall into the public domain.
None of these solutions is perfect. Moving forward, copyright law should focus on embracing a creative partnership between AI and humans, fostering a collaborative environment in which the strengths of both can be maximized. Humans bring creativity, intuition, emotional depth and contextual understanding to the artistic process. On the other hand, AI excels in processing vast amounts of data, identifying patterns, incorporating diverse fields of knowledge into creative outputs and offering new perspectives and possibilities.
Additionally, AI hallucination can be a good thing for artists, and copyright law should recognize this. AI hallucination refers to instances where AI generates information or content that is not based on reality or factual data. In professions like law, factual integrity is paramount, and AI hallucinations can lead to misinformation. But, unlike the legal field, art thrives on imagination, abstraction and the breaking of conventional boundaries. AI-generated hallucinations can serve as springboards for artistic creativity, pushing artists to think beyond traditional constraints. The unexpected and often surreal outputs can lead to innovative artworks that challenge norms and explore new artistic territories.
It is time to engage in these larger discussions so that the perception of artists using AI evolves. AI is not merely a tool like a brush, paper or camera; it is a creative partner that offers unprecedented possibilities, and it may be the most significant advancement since the last Industrial Revolution. I urge lawmakers and the Copyright Office to embrace a mentality of partnership between AI and humans when it comes to rulemaking.
Jiaying Jiang is an assistant professor of law at the University of Florida’s Levin College of Law. Her research focuses on policies and regulations regarding emerging technologies, including AI, fintech, blockchain, cryptocurrencies and central bank digital currencies. The author thanks Cai Studio for the demo of cAI™.
