Copyright vs. Creation | Who Really Owns Art in the Age of AI?

Imagine typing a few whimsical words into a text box: “An astronaut riding a unicorn on Mars in the style of Van Gogh.” Seconds later, a stunning, unique image appears on your screen, ready to be used. This is the magic of generative AI, a technology that has democratized creation on an unprecedented scale. But this new ease of creation has thrown a wrench into the centuries-old machinery of copyright law. Who is the true author of this digital masterpiece? Is it the person who wrote the prompt, the complex algorithm that generated the pixels, or the corporation that built and trained the AI? This article dives into the complex legal and ethical maze of AI art, exploring who really holds the rights to creation in our new technological age.

The human hand vs. the algorithm’s mind

At the heart of copyright law, particularly in jurisdictions like the United States, lies a foundational principle: the human authorship requirement. For a work to be eligible for copyright protection, it must be the product of human creativity. A photograph is protected because a person chose the subject, framing, and lighting. A novel is protected because a person crafted the plot and prose. The law was built to protect and incentivize human endeavor. AI art challenges this directly. When a user provides a simple prompt, can that be considered enough creative input to qualify as authorship?

The answer is murky and depends on the degree of human intervention. Consider the difference between these two scenarios:

• Scenario A: A user types “a golden retriever in a library” and uses the first image generated.
• Scenario B: An artist crafts a highly detailed, multi-paragraph prompt, specifying camera angles, lens types, lighting, character expressions, and composition. They then generate hundreds of images, selecting specific parts from several, and meticulously combine and edit them in a separate program to create a final composite piece.

In Scenario A, the user’s creative contribution is minimal. The AI did the heavy lifting, making countless “choices” about style and composition based on its training. Legal bodies like the US Copyright Office have been clear that such works lack the human authorship needed for protection. Scenario B, however, presents a stronger case. Here, the AI is used more like a sophisticated tool, akin to a camera or a paintbrush, with the final work being heavily guided and shaped by the human’s creative decisions. The key question courts and copyright offices are grappling with is defining where the line for “sufficient human authorship” truly lies.

The ghost in the machine: Can an AI be an author?

If the human prompter isn’t always the author, could the AI itself hold the copyright? The simple answer, under current law, is a firm no. Legal systems worldwide do not recognize non-humans as authors. An AI, no matter how sophisticated, lacks legal personhood, consciousness, and intent. It cannot enter into contracts, sue for infringement, or own property, including intellectual property. The most famous parallel is the “monkey selfie” case, where a court ruled that a macaque who took a picture of himself could not be the copyright holder because he was not a human.

Despite this legal reality, the philosophical debate continues. Proponents argue that generative AI is more than just a passive tool. It synthesizes concepts and creates novel compositions in ways that are not explicitly directed by the user. It makes millions of micro-decisions to produce the final image, demonstrating a form of emergent creativity. However, the counterargument is more legally sound. An AI is a complex statistical model that reflects and remixes the patterns present in its vast training data. It doesn’t understand what a unicorn or Van Gogh’s style is; it simply associates those words with pixel patterns it has been shown. It has no will of its own, making its “creation” a highly advanced form of mimicry, not authorship.

Training data: The original sin of AI art?

The conversation about AI authorship is inseparable from the controversial foundation upon which these models are built: their training data. Generative AI models like Midjourney and Stable Diffusion learn their skills by analyzing billions of images and text pairs, most scraped from the open internet. This dataset inevitably includes a colossal amount of copyrighted material, from photographs and illustrations to famous artworks, all typically used without the original creators’ permission or compensation.

AI companies argue this practice falls under fair use (in the US) or fair dealing (in other regions), contending that using the data for training is a transformative purpose. They claim they aren’t re-selling the images but are using them to teach a system, which is fundamentally different. Artists and creators, however, see it differently. They argue their work is being ingested and used to power a machine that can then replicate their style, devalue their skills, and potentially eliminate their jobs. This has led to a wave of high-profile lawsuits, with artists and stock photo agencies like Getty Images suing AI firms for mass copyright infringement. This issue is the ethical core of the debate. If an AI’s output is fundamentally derived from copyrighted works, can it ever be truly “original,” and who should benefit from its creations?

Navigating the new frontier: Current rulings and future paths

As the technology races ahead, the legal world is slowly trying to catch up. The US Copyright Office (USCO) has become a key battleground, offering early guidance on the issue. In a landmark decision involving the comic book Zarya of the Dawn, the USCO granted copyright for the text and the creative arrangement of the panels, which were created by a human. However, it explicitly refused to grant copyright for the individual AI-generated images within the comic, ruling they were not the product of human authorship.

This ruling provides a glimpse into the current path forward for creators: copyright may be available for works that incorporate AI elements, but only the human-authored components will be protected. To strengthen their claim, creators using AI are advised to:

• Engage in significant post-processing and editing of AI outputs.
• Combine AI-generated elements with their own original drawing, photography, or design.
• Keep detailed records of their creative process, demonstrating how their choices and skills shaped the final product.

Looking ahead, the legal framework must evolve. Potential solutions range from creating new legislation designed specifically for AI-generated works to establishing licensing systems where AI companies pay royalties for the data used in training. Some have even proposed a new, limited form of “AI-generated work” protection or having such works immediately enter the public domain. The road is uncertain, but one thing is clear: the status quo is no longer sufficient.

The dawn of AI art has blurred the lines between creator and tool, inspiration and infringement. We stand at a legal and ethical crossroads where the fundamental requirement of “human authorship” is being tested like never before. As we’ve seen, current laws are ill-equipped for this new paradigm, leaving the human prompter, the AI, and the original artists whose work trained the models in a complex limbo. Rulings like the one for Zarya of the Dawn suggest a hybrid approach, protecting human creativity while leaving purely machine-made outputs in the public domain. Ultimately, forging a fair future will demand a delicate balance: fostering technological innovation while fiercely protecting the rights, recognition, and livelihoods of human artists. The answer we find will redefine what it means to create art.

Image by: Pachon in Motion
https://www.pexels.com/@pachon-in-motion-426015731