7 OCTOBER | LONDON 2024
SEPTEMBER 12TH - 14TH
The O2, LONDON
Can AI-generated images be copyrighted?
By the CogX R&I team
Content creation with the help of computers is nothing new. What has changed is the extent to which machines can now contribute to our creative process.
However, as AI-generated content proliferates across various industries, the question of who—or what—deserves credit for these creations is becoming increasingly complex.
For example, if an artist performs significant creative work on an AI-generated piece, the human-generated portion could be seen as original and therefore eligible for copyright protection — but beyond that, things get murky.
A critical issue then becomes defining what exactly qualifies as "significant creative work". Is it the careful selection of elements? The way they’re arranged? Or perhaps it’s the application of filters and effects that give the piece its final polish?
Consider a scenario where you're working on a fantasy novel and want an image of a dragon perched on a mountain cliff at sunset for the cover. You use an AI tool like Midjourney to generate the scene. After a few prompts, the AI produces a stunning dragon, but the background isn't quite right. You decide to create a composite image to achieve your vision.
First, you generate an AI image of a dragon. Since this image is entirely AI-generated, with no human modifications, it isn’t eligible for copyright. Anyone could use it freely.
Next, you use Midjourney to generate a background of a fantasy setting with a dramatic sunset. Again, because this image is purely AI-generated, you may not be able to claim any copyright on it.
But If you import these AI-generated images into graphic design software like Photoshop, and start layering the dragon onto the cliff, adjusting its position, and blending the sunset to create a cohesive image, you’re beginning to add what the U.S. Copyright Office might consider "appreciable content". This means you’ve potentially done enough creative work — through your selection, arrangement, and manipulation of these elements — to secure a copyright for the final composite image.
However, even with these edits, there’s no guarantee that your work will be eligible for copyright protection.
So what can and can't be protected? Recent cases and emerging laws may provide some clarity on this complex legal issue.
Do AI-generated images have copyright protection?
When it comes to the legalities, the U.S. Copyright Office (USCO) has made its position crystal clear, works with significant AI input don’t qualify for copyright protection in the United States.
Recent guidelines state that AI-generated content exceeding a "de minimis" threshold must be identified and excluded from copyright registration. In plain terms, while human-created elements of AI-assisted works, like "prompt instructions", can still be copyrighted, anything generated by an AI system like DALL-E or Midjourney can’t be.
So why the strict stance?
It all boils down to how these generative AI systems operate. These models, like other machine learning systems, don't create out of thin air; they identify and replicate patterns from data they've been trained on. For instance, if an AI-generated painting channels the essence of Picasso, it’s not because the AI suddenly tapped into some hidden well of artistic genius. Rather, it’s because the system has been trained on a vast array of Picasso's works.
According to USCO guidelines, AI systems, even when used by humans, cannot be considered legal owners of their creations. This is especially pertinent given that much of the source material these systems use has been scraped from the internet, often without permission — and may already be protected by existing copyright laws.
The situation in the UK
The UK stands out as one of the few countries with a specific legal framework for computer-generated works.
The UK’s Copyright, Designs and Patents Act 1988 (CDPA) introduced Section 9(3), a provision that uniquely designates the person who makes the necessary arrangements for creating a work as its author, bypassing the traditional need for a human creator.
These works, known as LDMA works, enjoy protection for 50 years from the end of the calendar year in which they were created.
But even this provision is under the microscope, thanks to rapid advancements in AI. While the CDPA aims to foster AI innovation, it leaves a lot of questions unanswered — particularly around the concept of originality, a cornerstone of copyright protection. The big question here is “how do you apply originality to works generated by AI with no human input?”
Adding to the complexity, there’s also uncertainty about whether this 'human stand-in author' needs to meet the same originality standard required for traditional copyright.
This approach contrasts sharply with the more restrictive interpretations — like those from the US Copyright Office — which has outright denied copyright protection for AI-generated works.
The complexity of human-AI collaboration
When humans and machines collaborate on creative projects, things get even more complicated. Recent high-profile cases have brought these challenges into sharp focus.
In September 2023, the USCO made headlines by denying copyright protection for Jason Allen’s "Théâtre D'opéra Spatial”, which had won the top prize at the Colorado State Fair the previous year.
Allen, who used Midjourney to create the work, insisted that his creative input — crafting prompts, adjusting scenes, and selecting focus areas — made him the author. But the USCO disagreed, ruling that the piece "lacks human authorship" and thus falls outside the scope of copyright law, which "excludes works produced by non-humans”.
Another headline-grabbing case involved Kristina Kashtanova’s comic book, "Zarya of the Dawn." Initially, the USCO granted copyright protection for the work, created using Midjourney. But in a surprising twist, the office later partially withdrew this protection.
The USCO did recognise Kashtanova’s copyright in the text and the selection and arrangement of elements, but it denied protection for the individual AI-generated images, stating they were "not the product of human authorship".
The USCO Review Board acknowledged that copyright could apply to a work where a computer was used as an “assistive instrument,” but the line between human-directed authorship and mere AI assistance remains anything but clear.
What is the future of AI and copyright?
The challenges facing AI in the realm of copyright are far from over. If left unchecked, the use of copyrighted material in AI models could pose a significant threat, not only to artists but to the technology's future as well.
Courts around the world will play a pivotal role in determining the extent of protection and recognition granted to human input in AI-generated works. The path forward, however, will likely involve a delicate balance between fostering innovation and protecting intellectual property rights.
Popular Articles
Get the CogX Newsletter
Get the latest tech news in your inbox each week
Undoing the Tech Coup: A Thrilling Conversation with Marietje Schaake
Issue 37
Marietje Schaake,a former European Parliament member and Fellow at Stanford University’s Cyber Policy Center and the Stanford Institute for Human-Centered Artificial Intelligence, discusses the strategies outlined in her new book, 'The Tech Coup: How to Save Democracy from Silicon Valley,' on how to reclaim democratic control in the digital age.