Artificial Intelligence has thrown a major spanner into the commissioned artwork economy. For most artists, this change is far from welcome. The emergence of AI-generated art threatens to undermine the market demand for creative work, both from competing non-artists and employers who mistakenly assume that good art can now be created in seconds. Should artists leverage new technology to compete against these threats, and what legal boundaries should they look out for while they do so?
As of September 2023, there are no laws that outright prohibit the sale of AI-generated artwork. This means that anything generated by AI is free game, right? Well, not quite. Companies such as Getty Images as well as individual artists like Sarah Anderson and Kelly McKernan have filed lawsuits against AI companies on the basis of using copyrighted images for commercial purposes.
Because AI models process millions of images within their training data, there’s no foolproof way to tell whether a generated image is an amalgamation of other images or complete recreation of something that’s already out there. And even if the AI output is proven to be ‘original’, we still have to contend with the fact that AI models wouldn’t otherwise be capable of generating outputs with the same level of detail or accuracy had it not been for the existence of works that came before it in the first place.
But artists themselves are inspired by the works of others all the time, right? Well, yes. But the way that most artists use existing work to inspire their own often falls under the definition of fair use. In legal terms, fair use is a defence against copyright infringement. As long as a replication of copyright material falls into the categories of (1) commentary and criticism, or (2) parody, using another person’s work is completely legal.
However, the use cases for fair use are often distinct from how self-proclaimed “AI artists” attempt to monetise AI generated work. Commentary and criticism, for example, often quote a few lines of a book or song to make a point. In parody, significant enhancements are often made to the original work and clearly distinguishes it from the original.
In the case of AI-generated work, the case for fair use does not clearly fall into either category, and in fact clearly violates the fourth factor for measuring fair use: the effect upon the potential market. According to the Stanford Libraries, “depriving a copyright owner of income is very likely to trigger a lawsuit.”, which is in fact what has played out as artists who used to obtain commissions from clients requesting a particular style are now losing business to individuals who wish to take a shortcut by extracting similar images from AI models instead.
If AI generated content teeters along the legal lines of copyright, can it then be modified and later copyrighted for commercial purposes? As stated earlier, fair use does permit limited use of copyrighted material, in surprisingly precise measurements. Examples include:
With these measurements, however, it’s clear to see that raw AI-generated content will likely not qualify for fair use. In fact, the U.S. Copyright Office has long taken the stance that no copyright protection should be offered to work created by non-humans, including machines. This means that anyone who wishes to use AI tools for commercial purposes should be prepared to make significant amendments if they wish to stay on the safe side of the law.
This standard was tested in September 2022 when the U.S. Copyright Office historically granted copyright of the AI generated graphic novel Zarya of the Dawn to author Kris Kashtanova. However, this copyright was eventually revoked after it was decided that the novel’s characters, dialogue, and images — all of which were generated using Midjourney — were a result of “non-human authorship”. Only the text itself and the specific arrangement of written and visual elements, which required intervention from Kashtanova, remained protected. And in case anyone was in doubt, Kashtanova did in fact make changes to Midjourney’s generated outputs, but these changes were deemed to be too insignificant to necessitate copyright protection.
It should be noted that this ruling technically does not stop Kashtanova from profiting off Zarya of the Dawn. However, not being copyright protected makes such an endeavour less viable as Kashtanova would not be able to stop other creators from reusing her work for their own creative monetisation endeavours.
“In order for there to be infringement, there has to be an author. So, if there isn’t an author, I don’t know that there can be infringement.”
Some argue that if it is not the creators of training data nor the AI prompters who own AI generated content, then it should be the developers of AI models who should have ownership. But this argument still falls flat based on current legal parameters because machine learning models by definition identify and replicate patterns from existing data that is copyrighted in one way or another. It’s outputs are therefore simply a culmination of already existing work, rather than a new creation in its own right.
It’s important to note that AI art models offer varying levels of ownership. Some AI art generators such as DALL-E and Wombo allow you to monetise generated artwork, others like MidJourney’s free account do not offer the same level of ownership. However, do proceed with caution when using these tools as the legal basis for these ownership clauses still remain in question.
These laws will not entirely put the brakes on AI. Considering the speed at which AI content can be generated, there’s nothing stopping companies from shifting all the productions in-house or individual creators from selling AI-generated content under the radar, all of which will undermine individual artists.
Still, we believe that artists who are capable of producing quality work, understanding market demands, and communicating the value that they bring to clients will still rise above. After all, designing forms but a small part of the modern artist’s day to day job, which can involve everything from marketing strategy to project management and even sales conversion.
Both public and private companies have come out to issue company guidelines for the use of generative AI tools due to cybersecurity and data secrecy concerns. For example, public officers working in Singapore’s Ministry of Communications and Information (MCI) have been allowed to use ChatGPT for research and writing, but ultimately must take responsibility for copyright adherence and accuracy of any information that they publish.
Other companies such as Fantasy Franchise Dungeons & Dragons, have taken a more extreme approach to outright ban the use of AI altogether after prominent artists within the company have had their names used as prompts to generate art in their style. So when in doubt, do seek the permission from your specific client before using AI software. Otherwise, you risk having a lawsuit on your hands when proprietary images or information is extracted from prompts by the public.
We still might be a couple years away from full dive reality, but you don’t need to wait that long to start working on projects in virtual reality, mixed reality, or the metaverse. Early sign ups to the Freelancer Nation network will receive their very own NFT access pass and endless opportunities to experiment with and showcase their best work on Web3 and other platforms of the future.
To stay up to date with the latest developments on the creative scene, subscribe to our Freelancer Nation newsletter for monthly updates on the latest social media trends and workshops specially curated to help creative freelancers supercharge their careers.
Subscribe to Freelancer Nation!
