By: Jacob Alhadeff

Copyright has insufficiently analyzed causation for both authorship and liability because, until now, causation was relatively obvious. If someone creates a painting, then they caused the work and receive authorial rights. If it turned out that the painting was of Mickey Mouse, then that painter may be liable for an infringing reproduction. However, recent technological advances have challenged the element of causation in both authorship and infringement. In response, recent law and scholarship have begun to address these issues. However, because they have addressed causation in isolation, current analysis has provided logically or ethically insufficient answers. In other words, authorial causation has ignored potential implications for an entity’s infringement liability, and vice-versa. Regardless of how the law responds, generative AI will require copyright to explore and enumerate the previously assumed causation analyses for both infringement and authorship. This blog explores how generative AI exposes the logical inconsistencies that result from analyzing authorial causation without analyzing causation for infringing reproductions.

Generative AI largely requires the following process: (1) an original artist creates works, (2) a developer trains an AI model on these works, and (3) an end-user prompts the AI to generate an output, such as “a mouse in the style of Walt Disney.” This generative AI process presents a novel challenge for copyright in determining who or what caused the output because generative AI challenges conventional notions of creation.

Causing Infringement

Andersen et al. recently filed a complaint against Stability AI, one of the most popular text-to-art foundation models. This class action alleges that Stability AI is directly liable for infringing that result from end-user prompted generations. However, in a recent decision more closely analyzing causation and volition in infringement, the Ninth Circuit found that “direct liability must be premised on conduct that can reasonably be described as the direct cause of infringement.” Stability AI should not be found directly liable for infringing these artists’ copyright, in part because Stability AI cannot reasonably be said to be the direct cause of infringement. Such a finding would be similar to holding Google liable for reproducing images of Mickey Mouse on people’s computer screens when they search for “Mickey Mouse.”  

This lawsuit is particularly relevant since end-users have prompted thousands of generations that include the phrase “Mickey Mouse” and many appear substantially similar to Disney’s Mickey. If thousands of end-users have intentionally prompted the AI to generate Mickey Mouse, then what volitional conduct can most reasonably be described as the direct cause of infringement? It is clearly the end-user. However, what if the end-user simply prompted “a cartoon mouse” and the AI generated an infringing image of Mickey? Here, the end-user may not have intended to generate Mickey and reasonable notions of fairness may not find the end-user as the most direct cause of infringement. However, copyright is a strict liability tort, meaning that liability attaches regardless of a reproducer’s intent. Therefore, unless copyright applies an intentional or a negligence theory for direct liability, which it should not, then whomever or whatever is liable for infringing outputs shall be liable for both of the infringing outputs— “Mickey Mouse” and “a cartoon mouse.” Such an outcome not only feels deeply unfair, but it is unreasonable to say that the end-user is the direct cause of infringement when prompting “a cartoon mouse,” and vice versa. 

Cases called to answer similar questions have recently grappled with these same issues of volition and causation. Generally, courts have been hesitant to find companies liable for actions that are not reasonably deemed volitional conduct causing infringement. The court in Cartoon Network, for example, found that “volition is an important element of direct liability.” In the Loopnet case, the court found that “the Copyright Act… requires conduct by a person who causes in some meaningful way an infringement.” In this way, the law has so far mirrored our prior intuitions of fairness. Legal scholarship has noted that when copyright law has grappled with novel technology, it has found that causation in infringement requires volition that “can never be satisfied by machines.” This reasoning, as applied  to generative AI, may mean that an AI company should not normally be directly liable for outputs that infringe the reproduction right. 

Causing Authorship

This causation analysis has also begun for authorship rights. One copyright scholar compellingly argues that copyright law should explicitly enumerate a causal analysis for granting authorship rights. Such an analysis would follow tort law’s two step causation analysis including: (1) creation in fact and (2) legal creation. Aviv Gaon surveys authorial options in The Future of Copyright in the Age of AI, writing that there are those that favor assigning authorship to the end-user prompter, the AI developer, finding outputs joint works, or even attributing authorship to the AI itself. The simplest legal option would be to treat AI like a tool and grant authorship to the end-user. This is exactly how the law responded when photography challenged conventional notions of creativity and authorship. Opponents of finding photographers as authors argued that photography was “merely mechanical, with no place for… originality.” The Supreme Court in Burrow Giles instead found that the photographer “gives effect to the idea” and is the work’s “mastermind” deserving of copyright. 

However, treating AI like a conventional tool is an inconsistent oversimplification in the current context. Not only is it often less analogous to say that an end-user prompter is the ‘mastermind’ of the output, but AI presents a more attenuated causation analysis that should not result in  a copyright for all AI-generations. As an extreme example, recent AIs are employing other AIs as replicable agents. In these circumstances, a single prompt could catalyze one AI to automatically employ other AI agents to generate numerous potentially creative or infringing outputs. Here, the most closely linked human input would be a prompt that could not be said to have masterminded or caused the many resultant expressive outputs. Under Balganesh’s framework, no human could reasonably be found as the factual or legal cause of the output. Such use-cases will further challenge the law’s notions of foreseeability as reasonable causation becomes increasingly attenuated.

Importantly, in the face of this ongoing debate and scholarship, the Copyright Office recently made their determination on authorship for AI-generated works. In February 2023, the US Copyright Office amended its decision regarding Kristina Kashtanova’s comic book, Zarya of the Dawn, stating that the exclusively AI-generated content is not copyrightable.  Ms. Kashtanova created her comic book using Midjourney, a text-to-art AI, to generate much of the visual art involved. The copyright office stated that her “selection, coordination, and arrangement” of AI-generated images are copyrightable, but not the images themselves. The Office’s decision means that all exclusively AI-generated content, like natural phenomena, is not the type of content copyright protects and is freely accessible to all. The Office’s decision was based on their interpretation that “it was Midjourney—not Kashtanova—that originated the ‘traditional elements of authorship.’” The Office’s decision is appropriate policy, but when analyzed in conjunction with the current law on causation in infringement, it is inconsistent and may result in an asymmetrical allocation of the rights and duties that attend creation. Relevantly, how can a machine that is incapable of volition originate art? This is one of many ontological paradoxes that AI will present to law. 

Symmetrically Analyzing Causation

Two things are apparent. First, there is a beautiful symmetry in AI-generations being uncopyrightable, and the machines originating such works symmetrically do not have sufficient volition to infringe. If such a system persists, then copyright law may not play a major role in generative AI, though this is doubtful. Second, such inconsistencies inevitably result from causation analyses for mechanically analogous actions that only analyze one of infringement or authorship. Instead, I propose that copyright law symmetrically analyze mechanically analogous causation for both authorship and infringement of the reproduction right. Since copyright law has only recently begun analyzing causation, it is reasonable, and potentially desirable, that the law does not require this symmetrical causation. After all, the elements of authorship and infringement are usefully different. However, what has been consistent throughout copyright is that when an author creates, they risk both an infringing reproduction and the benefits of authorship rights. In other words, by painting, a painter may create a valuable copyrightable work, but they also may paint an infringing reproduction of Mickey Mouse. Asymmetrical causation for AI art could be analogized to the painter receiving authorship rights while the company that made the paintbrush being liable for the painter’s infringing reproductions. Such a result would not incentivize a painter to avoid infringement, and thereby improperly balance the risks and benefits of creation. Ultimately, if the law decides either the end-user or the AI company is the author, then the other entity should not be asymmetrically liable for infringing reproductions. Otherwise, the result will be ethically and logically inconsistent. After all, as Antony Honore wrote in Responsibility and Fault, in our outcome-based society and legal system, we receive potential benefit from and are responsible for the harms reasonably connected to our actions.

By: Jacob Alhadeff

In the early 2000s, courts determined that the emerging technology of peer-to-peer “file-sharing” was massively infringing and categorically abolished its use. Here, the Ninth Circuit and Supreme Court found that Napster, Aimster, and Grokster were secondarily liable for the reproductions of their users. Each of these companies facilitated or instructed their users on how to share verbatim copies of media files with millions of other people online. In this nascent internet, users were able to download each other’s music and movies virtually for free. In response, the courts held these companies liable for the infringements of their users. In so doing, they functionally destroyed that form of peer-to-peer “file-sharing.” File-sharing and AI are in not analogous, but multiple recent lawsuits present a similarly existential question for AI art companies. Courts should not find AI art companies massively infringing and risk fundamentally undermining these text-to-art AIs.

Text-to-art AI, aka generative art or AI art, allows users to type in a simple phrase, such as “a happy lawyer,” and the AI will generate a nightmarish representation of this law student’s desired future. 

Currently, this AI art functions only because (1) billions of original human authors throughout history have created art that has been posted online, (2) companies such as Stability AI (“Stable Diffusion”) or Open AI (“Dall-E”) download/copy these images to train their AI, and (3) end-users prompt the AI, which then generates an image that corresponds to the input text. Due to the large data requirements, all three of these steps are necessary for the technology, and finding either the second or third steps generally infringing poses and existential threat to AI Art. 

In a recent class action filed against Stability AI, et al (“Stable Diffusion”), plaintiffs allege that Stable Diffusion directly and vicariously infringed on the artist’s copyright through both the training of the AI and the generation of derivative images, i.e., steps 2 and 3 above. Answering each of these claims requires complex legal analyses. However, functionally, a finding of infringement on any of these counts threatens to fundamentally undermine the viability of text-to-art AI technology. Therefore, regardless of the legal analysis (which likely points in the same direction anyways) courts should not find Stable Diffusion liable for infringement because doing so would contravene the constitutionally enumerated purpose of copyright—to incentivize the progress of the arts. 

In general, artists have potential copyright infringement claims against AI Art companies (1) for downloading their art to train their AI and (2) for the AI’s substantially similar generations that the end-user prompts. In the conventional text-to-art AI context, these AI art companies should not be found liable for infringement in either instance because doing so would undermine the progress of the arts. However, a finding of non-infringement leaves conventional artists with unaddressed cognizable harms. Neither of these two potential outcomes are ideal. 

How courts answer these questions will shape how AI art and artists function in this brave new world of artistry. However, copyright infringement, the primary mode of redress that copyright protection offers, does not effectively balance the interests of the primary stakeholders. Instead of relying on the courts, Congress should create an AI Copyright Act that protects conventional artistry, ensures AI Art’s viability, and curbs its greatest harms. 

Finding AI Art Infringing Would Undermine the Underlying Technology

A finding of infringement for the underlying training or the outputs undermines AI Art for many reasons: copyright’s large statutory damages, the low bar for granting someone a copyright, that works are retroactively copyrightable, the length of copyright, and the volume of images the AI generates and needs for training.

First, copyright provides statutory damages of $750 to $30,000 and up to $150,000 if the infringement is willful. Determining the statutory value of each infringement is likely moot because of the massive volume of potential infringements. Moreover, it is likely that if infringement is found, AI art companies would be enjoined from functioning, as occurred in the “file-sharing” cases of the early 2000s. 

Second, the threshold for a copyrightable work is incredibly low, so it is likely that many of the billions of images used in Stable Diffusion’s training data are copyrightable. In Feist, the Supreme Court wrote, “the requisite level of creativity is extremely low [to receive copyright]; even a slight amount will suffice. The vast majority of works make the grade quite easily.” This incredibly low bar means that each of us likely creates several copyrightable works every day. 

Third, works are retroactively copyrightable, meaning that the law does not require the plaintiff to have registered their work with the copyright office to receive their exclusive monopoly. Therefore, an author can register their copyright after they are made aware of an infringement and still have a valid claim. If these companies were found liable, then anyone with a marginally creative image in a training set would have a potentially valid claim against a generative art company.

Fourth, the copyright monopoly lasts for 70 years after the death of the author. Therefore, many of the copyrights in the training set have not lapsed. Retroactive copyright registration combined with the extensive duration of copyrightability means that few of the training images are likely in the public domain. In other words, “virtually all datasets that will be created for ML [Machine Learning] will contain copyrighted materials.”

Finally, as discussed earlier, the two bases for infringement claims against the AI art companies are (1) copying to train the AI and (2) copying in the resultant end generation. Each basis would likely result in billions or millions of potential claims, respectively. First, Stable Diffusion is trained on approximately 5.85 billion images which they downloaded from the internet. Given these four characteristics of copyright, it is likely that if infringement were found, many or all of the copyright owners of these images would then have a claim against AI art companies. Second, regarding infringement of end generations, Dall-E has suggested that their AI produces millions of generations every day. If AI art companies were found liable for infringing outputs, then any generation that was found to be substantially similar to an artist’s copyrighted original would be the basis of another claim against Dall-E. This would open them up to innumerable infringement claims every day. 

At the same time, generative art is highly non-deterministic, meaning that, on its face, it is hard to know what the AI will generate before it is generated. The AI’s emergent properties, combined with the subjective and fact-specific “substantial similarity” analysis of infringement, do not lend themselves to an AI Art company ensuring that end-generations are non-infringing. More simply, from a technical perspective, it would be near-impossible for an AI art company to guarantee that their generations do not infringe on another’s work. 

Finding AI art companies liable for infringement may open them up to trillions of dollars in potential copyright lawsuits or they may simply be enjoined from functioning.

An AI Copyright Act

Instead, Congress should create an AI Copyright Act. Technology forcing a reevaluation of copyright law is not new. In 1998, Congress passed the DMCA (Digital Millennium Copyright Act) to fulfill their WIPO (World Intellectual Property Organization) treaty obligations, reduce piracy, and facilitate e-commerce. While the DMCA’s overly broad application may have stifled research and free speech, it does provide an example of Congress recognizing copyright’s limitations in addressing technological change and responding legislatively. What was true in 1998 is true today. 

Finding infringement for a necessary aspect of text-to-art AI may fundamentally undermine the technology and run counter to the constitutionally enumerated purpose of copyright—“to promote the progress of science and useful arts.” On the other hand, finding no infringement leaves these cognizably harmed artists without remedy. Therefore, Congress should enact an AI Copyright Act that balances the interests of conventional artists, technological development, and the public. This legislation should aim to curb the greatest harms posed by text-to-art AI through a safe harbor system like that in the DMCA.