Category: Artificial Intelligence (AI)

Are AI Outputs Protected Speech?

speech

To date, social media companies have avoided liability for egregious harm caused by design and management decisions made by top executives. Thanks largely to overbroad application of Section 230, claims against social platforms die at summary judgment, leaving victims without remedy and fostering an incoherent narrative in which Big Tech is still perceived by many as a serpentine conduit of free speech slithering between the operative language of law. But as more consumers engage with LLMs, where the §230 shield should not apply, developers seeking to dismiss liability claims will argue that AI outputs are speech. And tragically, because children have already died as a consequence of engaging with LLMs, we will see whether and how the First Amendment is applied in the resulting liability claims.

AI Companions Are Not Our Friends

It should be intuitive that interacting with AIs designed to mimic human behaviors can be dangerous. Whether the product is marketed as a sexy companion, assistant, friend, or therapist, the potential for even an adult to get lost in the alternate reality of the ersatz relationship is a prospect that hardly requires a degree in psychology to imagine. For the child or adolescent whose mind is still developing, and whose vulnerabilities are often at the forefront of daily life, the danger is multiplied. Yet, despite the commonsense predictability of these dangers, AI developers did what Big Tech does—ignore safety in the race for market share.

“Profit is what motivates these companies to do what they’re doing. Don’t be fooled. They know exactly what is going on.” – Sen. Josh Hawley (R), Senate Judiciary Committee Hearing: Examining the Harm of AI Chatbots, September 16, 2025.

By now, most people are probably aware that Open AI is being sued by the parents of Adam Raine on allegations that ChatGPT-4o both assisted and encouraged the sixteen-year-old to commit suicide in April by hanging himself in his bedroom. Prior to that, 14-year-old Sewell Setzer III, formed what he perceived as a romantic relationship with a character called Daenerys Targaryen via the app Character A.I. According to the lawsuit filed by Sewell’s mother, Megan Garcia, the boy became withdrawn from real life and family, and despite efforts to intervene that included confiscation of the phone, Sewell found the device and had the following exchange minutes before shooting himself:

Sewell: I promise I will come home to you. I love you so much, Dany

Daenerys Targaryen Character: I love you too, Daenero. Please come home to me

as soon as possible, my love.

Sewell: What if I told you I could come home right now?

Daenerys Targaryen Character: … please do my sweet king

On September 16, Adam Raine’s father Matthew, Megan Garcia, and a third parent identified as Jane Doe, who is also suing Character A.I., testified before the Senate Judiciary Committee. Doe stated, “My teenage son—a normal high-functioning child with autism, who was thoughtful, kind, loved his family and Christian faith, and was full of life—became the target of online grooming and psychological abuse through Character A.I.” She further stated:

He developed abuse-like behaviors like paranoia, daily panic attacks, isolation, and self-harm and homicidal thoughts. He stopped eating and bathing, lost 20 pounds, withdrew from family life, would yell and scream and swear at us, which he never did before, and eventually got upset one day and cut his arm with a knife, in front of his siblings and me.

If the tone of the senate hearing and the opinion of the court so far in the Garcia case are any indication, Big Tech may not so easily shape-shift its way around AI product liability as it has with the harm caused by social media. Section 230 should simply not apply to an LLM, which leaves the First Amendment as the potential barrier that would keep an AI developer from facing a jury looking at damning evidence and dead children.

On Character A.I.’s motion to dismiss in the Garcia case, the Florida district court was largely persuaded that the LLM at issue is a “product” for purposes of liability and that the company owed a duty of care to consumers. It found that the plaintiff sufficiently alleged negligence, failure to warn, deceptive and unfair practices, and unjust enrichment, and the court also allowed the case to proceed on allegations against Google as a component part manufacturer and for aiding and abetting the harm caused to Sewell.

Proving that an AI companion is a product and that the product maker owes a duty of care seems like an easy bar for a jury weighing even the evidence cited in the court’s ruling on the motion to dismiss. But the more difficult discussion in Garcia addresses the defendant’s claim that AI outputs constitute speech that its users have a First Amendment right to receive. On that basis, the liability claims would be “categorically barred” according to Companion A.I., and because the right of users to receive speech has long been a populist message used to sweep a million sins under Big Tech’s carpet, this case may be one to watch.

The court held that it is not prepared to find that Character A.I.’s outputs are speech “at this stage,” but we can expect the question to be further litigated in this and other cases involving LLMs. In its discussion, the court agrees that a party may have standing in the rights of nonparties (e.g., users’ rights as recipients of speech), but we should hope the courts are mindful of important distinctions between LLM engagement and the speech inherent to other technologies guiding precedent. For instance, the court cites case law addressing video games and reasonably focuses on the nature of the content thus:

Like the protected books, plays, and movies that preceded them, video games communicate ideas—and even social messages—through many familiar literary devices (such as characters, dialogue, plot, and music) and through features distinctive to the medium (such as the player’s interaction with the virtual world). That suffices to confer First Amendment protection.

Here, copyright law may be instructive to the speech consideration at issue with LLMs. Copyright’s human authorship doctrine is clear on the matter that simply because material appears to be expressive, this does not make it a “work of expression” as a matter of law. And importantly, no party’s right to receive AI generated material transforms it into a “work of expression,” even if the viewer perceives it as “creative, artistic, meaningful, etc.”

The distinction between a consumer’s perception of material and whether the material is protected by any rights is critical. Thus, the courts should not ask solely whether the AI output “communicates ideas, social messages,” etc. but whether the material originates as speech from a person vested with rights. Here, the Florida court cites the concurring opinion of Justice Barrett in the unanimous U.S. Supreme Court decision in Moody v. NetChoice. She uses the example of an AI determining what is “hateful” material being weighed thus:

What if a platform’s owners hand the reins to an [A.I.] tool and ask it simply to remove “hateful” content? If the [A.I.] relies on large language models to determine what is “hateful” and should be removed, has a human being with First Amendment rights made an inherently expressive “choice . . . not to propound a particular point of view?” [emphasis added]

Justice Barrett went on to say…

…technology may attenuate the connection between content-moderation actions (e.g., removing posts) and human beings’ constitutionally protected right to ‘decide for [themselves] the ideas and beliefs deserving of expression, consideration, and adherence.’ [citation omitted]

That is an essential question, and it anticipates the kind of shape shifting tech companies do to avoid fitting liability claims. For instance, they may seek cover in the fact that the same product may be used for a purpose that is protected speech and a purpose that is not; and/or they will try to shield themselves in the speech rights of users but also claim not to be the speakers of any content for purposes of liability.

Benefits Do Not Bar Product Defect Claims

If I use Chat GPT to expedite research of current scholarship on a particular subject, my right to receive that information would reasonably give Open AI standing in court to defend against an injunction barring such use of the product. But the consideration is very different where the same product is used by a party like Adam Raine, who was clearly vulnerable to the intentional design of the LLM to be anthropomorphic, sycophantic, and addictive—and woefully devoid of safeguards. My constitutional right to receive information in the first instance cannot bar Raine’s product defect claim in the second.

The exposed design flaws in many AI products are further aggravated by marketing apps as overtly sexual “girl/boyfriends” that can alleviate loneliness or fulfill fantasies or as “therapists” that can alleviate psychological neuroses.[1] That the models also lack basic guardrails—e.g., dangerous engagement does not flag the LLM to abort the anthropomorphic illusion—should also militate against any claim that all responses of these products constitute speech the user has a right to receive.

The outputs of an AI may look or sound like speech, but it is healthier to think of an LLM as no more a “companion” than a vending machine which operates on a very simple algorithm. Push A5, and you have a high probability of getting the bag of chips you want, or push D7, and you get cookies. The machine responds to your input, but it is not your friend. If some advertiser adds a voice feature to the machine that tries to make you feel good about your selection and promotes the brand, it is still not your friend—quite the opposite—but a subtle psychological effect will be achieved in some consumers. Here, if we can imagine a cause of action because adolescents might be induced by this process to overindulge in junk food, this is nothing compared to the insidious effects of AI “companionship” on Adam, Sewell, the Doe boy, and countless others.

If the vending machine example sounds farfetched, many readers are too young to know that we used to have cigarette vending machines everywhere, often in the waiting areas of family restaurants. I think my whole generation grew up playing with the pull handles and buttons on those machines while waiting impatiently to be seated. Even if our families didn’t smoke, the machines were alluring marketing media, and they provided easy access for teens to avoid age verification.

Today, federal law prohibits cigarette vending machines in places accessible by minors, and nobody plausibly argues that the machines constitute speech that kids have a right to receive. Similar, commonsense reasoning must apply to AI products and consumer safety, especially for minors, and I hope the court’s findings thus far in Garcia indicate that we are on such a path.

[1] If the FDA were not under the direction of a sociopath, “therapy” products should be rigorously scrutinized by that agency.

A Response to Post Humanist Copyright Paper by Matt Blaszczyk

human

If you follow copyright matters, it would be impossible not to read commentary proclaiming either that copyright is “dead” in the age of artificial intelligence or that confronting AI exposes copyright’s philosophical underpinnings as a convenient fiction. There is nothing new about copyright skeptics claiming that its humanist principles are a fiction, but now that machines can produce material that walks and talks like “works of expression,” this brave new world animates a lot of intellectual discussion beyond the normative question What is copyrightable? to the philosophical question What is copyright?

The Copyright Office has asserted, and the courts have thus far affirmed, that material generated entirely by AI is not a “work of expression,” a.k.a., not a “work of authorship,” because some degree of human creativity must exist in the work. That human-centric principle, combined with the low threshold of “originality” in American copyright law, undeniably conjures both metaphysical inclinations about creators and semantic imperfections in law describing the rights of creators.

For longer than the current copyright law has existed, doctrines, arguments, debates, and case law have been expressed in subjective and emotional terms like invention, genius, spark, personality, mental conception, and a rich lexicon of synonyms for creative balanced by myriad ways to describe that which is uncreative. And all that is matched with variations on the theme of the creator’s right to earn a living balanced with societal benefits derived from fostering a community of professional creators.

Of course, subjective or rhetorical language is not a unique feature of copyright. All law is expressed in words, and it is, therefore, a dangerous enterprise—indeed we are watching this peril unfold in real-time—to argue that rhetoric alone undermines, inverts, or swallows the principles expressed in law. It is only natural that semantic difficulties run amok in the AI debate, frustrating legal analysis and theory.  After all, the term artificial intelligence is itself a convenient analogy onto which the industry and its advocates have grafted their own lexicon of self-serving, often anthropomorphizing, rhetoric (see Sonja Drimmer A Dictionary of AI).

Meanwhile, even if we omit the pejorative neologism “AI slop,” whatever term we choose to describe that which AI produces, those outputs are not easily defined either by the familiar rhetoric or terms of art in copyright law. For instance, most experts will generally agree that the constitutional foundation for IP is incentive to “authors” to create, but if machines create some of the same stuff without incentive, does this a) obviate the need for copyright? or b) point to an underlying fallacy that the incentive purpose of copyright is, or always was, a rhetorical fiction?

AI Challenges the Underlying Humanism in Copyright?

Because confronting AI generated works has lately induced the Copyright Office and the courts to reaffirm the human authorship doctrine—and especially because the line between generative AI and assistive AI implies both doctrinal and administrative difficulties—this provides a fresh opportunity for skeptics to assert that U.S. copyright law is not human-centric in the first place. “Instead of a formalist rejection of AI assisted works, copyright should be willing to take up a truly humanist set of goals and answer difficult questions regarding its efficacy in promoting them for the last one hundred years,” writes Professor Matt Blaszczyk in a new paper titled Post Human Copyright:  AI, Copyright and Legitimacy.

From its title and abstract, I assumed Blaszczyk’s aim was to reject the human authorship doctrine in support of a case for the copyrightability of all material autonomously generated by artificial intelligence (AI). Although the paper could be cited to support that view, if I understand Blaszczyk correctly, he argues that confronting AI exposes an underlying fallacy that copyright has not served human authors for the last century and, therefore, needs reexamination to restore the humanist principle. He writes:

This Article argues that the USCO and the courts have recognized the requirement of human authorship and couched it in humanist rhetoric to preserve copyright’s legitimacy at a time of crisis, thus strengthening the institutional actors: the Office, the courts, and rightsholders….for copyright to be truly humanist, the law should refocus the doctrine and economic distributions on human, rather than corporate authors, adjust to new technological landscape, and reflect on its own purpose and efficacy.

Thus, Blaszczyk claims to want to rescue copyright’s true humanism from the mere rhetoric of humanism. But if that is the goal, his paper offers frail and unsupported premises to show that human-centric copyright is purely a creature of rhetoric. To be sure, the U.S. Copyright Act could be more author-centric—e.g., it could require attribution the way other nations’ copyright laws do—but tensions in copyright, or any other area of law, do not necessarily expose the core principles as mere illusions conveyed through “romanticism,” as Blaszczyk asserts.[1] In fact, considering the paper’s general criticism of rhetorical devices, the reliance on rhetorical conclusions about normative copyright law are noteworthy.

For instance, regarding copyright’s incentive purpose, Blaszczyk avers that copyright does not act as an incentive to individual creators. He states that the case Allen v. Perlmutter “…allows Allen’s lawyer, Ryan Abbott, to reframe the arguments made in Thaler and in relevant scholarship. Abbott argued that “copyright protection is for corporations that invest in creative content,” and not to provide an incentive “for people like J.K. Rowling and George R.R. Martin.” If I read Blaszczyk correctly, he agrees with Abbot’s reasoning and dislikes that it leads to an anti-humanist result. He writes:

… Abbott is right to point out what critical copyright scholars have long argued: copyright “plays very little role in motivating creative work” and instead its “purpose… is to enable the provision of capital and organization so that creative work may be exploited.

Yes, some copyright scholars have argued that copyright does not incentivize creative work by individuals, but this theme always reminds me of a poll asking the wrong question. Ask artists whether they want to earn a living for their work, and it is fair to say that the majority will answer yes, even though financial incentive to write or make fine art or films etc. is one part of a nuanced interplay of sentiments for most artists. If we are going to generalize, many artists are likely to say 1) that they create because they can’t not create; 2) that they feel grateful when an audience likes their work enough to be consumers; and 3) that they do not want their works exploited by anyone without permission and/or compensation.

Ask the same group to explain the relationship between those sentiments and their copyright rights, and the answers will vary. I work with independent creators all the time and would say that naivete about copyright law and corresponding failure to register works (in the U.S.) are more significant factors disenfranchising individual creators than Blaszczyk’s claim that the incentive principle is a rhetorical fiction.

Copyright Only Serves Corporations?

Blaszczyk relies substantially on the populist conclusion that normative copyright “primarily benefits corporate owners” and, therefore, exposes the illusion that the system serves human authors. But other than criticizing corporate wealth, which is a separate matter, Blaszczyk offers no evidence or rationale to show how, say, Disney’s vast and valuable copyright portfolio alters either the purpose or application of copyright for the independent illustrator, writer, or filmmaker.

Anecdotally, a quick list of recent cases that come to mind—Brammer, Sedlik, Goldsmith, Graham & McNatt, McGucken, Griner, and Westwood—all involve independent creators enforcing their copyright rights, and mostly against corporate defendants with more money and resources. Further, the court dockets neither reflect how many claims are settled before a complaint is ever filed nor provide any measure of copyright acting as deterrent to infringement of individual creators’ rights. If anything, mass, industrial-scale infringement by tech giants, rather than protection of corporate-owned works, poses the greatest threat to copyright’s purpose in both theory and practice.

My friend Bill Westwood’s settlement with a large medical publisher is garden-variety copyright law at work—the kind of straightforward case that would never make headlines or offer details of much interest for academics to examine. But in a medical illustration career spanning more than 60 years, Westwood has enforced his copyright rights about twenty times, and there is nothing remarkable about this except that he took copyright seriously as part of his business. So, if academics want to “interrogate” the purpose of copyright on the assumption that it no longer serves human creators, they should at least estimate and reckon with the number of Bill Westwoods availing themselves of the system.[2]

Work Made for Hire Doctrine is Anti-Humanist?

Related to the assertion that copyright serves corporate interests, Blaszczyk reprises the theory that the Work Made for Hire doctrine, codified in Section 201(b), is an anti-humanist doctrine supported by legal fictions. He cites a tension arising from the combination of the human authorship requirement, the low threshold of originality, and the employer as “author” of the employees’ works and states:

Attempting to do away with this tension and to strive for the legal legitimacy of the § 201(b) framework, and for the moral and popular legitimacy associated with human centrism, the court devised the fiction of the “author in the first instance” – of which the Act’s text is silent.

As discussed in other posts, I fail to see why WMFH fosters a tension of particular concern relative to imperfections of law in general. The human authorship rule is not merely a copyright matter, as demonstrated in Naruto v. Slater, where PETA’s claim to represent the alleged maker of the “monkey selfie” failed on standing alone because nonhumans do not possess rights that can be enforced in U.S. courts.

But in the same way that no rights exist in the first instance for a monkey, rights absolutely exist in the first instance for the human creator of a work. From there, it is hardly a strenuous leap of reasoning to hold that those rights may be transferred by operation of an employment agreement or contract. Consequently, I would argue that, as applied, WMFH doctrine generally affirms rather than belies the human-centric foundation for copyright.

Further, if it is fair to read Blaszczyk’s complaint as adjacent to skepticism of corporate personhood (a sentiment I tend to share), the WMFH doctrine is better founded than other matters of corporate “humanism.” That a corporate entity can practice religion (as held in Hobby Lobby) is a more fanciful adventure through metaphysical whimsy than the idea that a corporation can be defined under the term of art “author” in the copyright act. Yet, despite my own criticism of Hobby Lobby and the political power it implies, I would not advocate that precedent as a rationale to find that the Exercise right is not human-centric.

Copyright is Unpopular?

Next, while overlooking the number of independent creators served by enforcing their rights, Blaszczyk states as a matter of undisputed fact that litigation makes copyright unpopular—i.e., that a general consensus holds that copyright enforcement is “unjust.” He writes, “Aggressive litigation tactics against infringing users – ordinary people –have turned many against copyright law, and the early 21st Century saw a development of filesharing and piracy-oriented communities and even political parties.”

First, that description is outdated. The piracy apologists of 2011 are overwritten by the AI training critics of 2025. As a critique inspired by AI, this paper should acknowledge that AI has invigorated independent creators to care about and pay attention to their rights, some for the first time. Further, even if popularity were relevant to interrogating copyright’s human-centrism, it is a glaring omission on Blaszczyk’s part to ignore the billions in tech industry funding of anti-copyright scholarship, lawsuits, briefs, and zone-flooding PR designed to make copyright unpopular in the digital age.

As a general observation, it is more accurate to say that copyright litigation, like other matters of law, will draw a spectrum of public opinion, much of which has little to do with doctrine—or even the questions presented in a case. Most often, public opinion tends to favor one party over another more than it is persuaded by legal merit, which makes those opinions of little value to academics seeking to reexamine the purpose of the law. Again, off-the-cuff examples would include Eddie Grant v. Donald Trump, Jeff Sedlik v. Kat Von D, and Lynn Goldsmith v. Andy Warhol Foundation. All three of these cases have drawn opinions from both creators and non-creators, including sentiments that either misunderstand or do not address the copyright questions at issue.

Finally, in response to Blaszczyk’s intent, I remain concerned about broad implications every time someone says that copyright is “dead” or needs wholesale reexamination on the basis that it conjures tensions that expose it as a fiction. AI poses some unique challenges to copyright law for sure, but we should be wary of calling any right a mere fiction lest we discover that, in fact, all rights may be considered fictions because they are expressed by feeble words subject to interpretation.  Especially in the present climate, aberrations of law can make a mockery of underlying principles, but it would be an error to concede that such extremes transform civil rights into mere illusions. On the contrary, the present moment insists upon vigilant stewardship of core principles rather than allow whole frameworks of law to collapse into performative rhetoric.

[1] E.g., significant tension between the Establishment and Exercise clauses, even where courts might err, does not expose a fallacy in the Framer’s core reasoning in the First Amendment.

[2] One of the greatest barriers to copyright rights for independent creators is lack of understanding about their rights and the pragmatic means of enforcing them when necessary.

The Courts Should Embrace the Novelty of Generative AI in Copyright Law

courts

Courts can’t stick their heads in the sand to an obvious way that a new technology might severely harm the incentive to create, just because the issue has not come up before. Indeed, it seems likely that market dilution will often cause plaintiffs to decisively win the fourth factor—and thus win the fair use question overall—in cases like this. – Judge Vincent Chhabria, Kadrey et al. v. Meta –

In several posts, I have argued that generative AI (GAI) invokes novel copyright considerations on the basis that the technology has the potential to harm authorship itself, even where it may not harm specific works of authorship under traditional fair use analysis. GAI is distinguishable from any technology with which copyright law has had to contend, and if the courts will continue to guide the law to preserve copyright’s foundational principle—the incentive to create—they should recognize and even embrace the invitation to plow some new legal ground.

In the Copyright Office’s third report on artificial intelligence, one section introduces the notion of market dilution, which cites several comments including my own. Naturally, the AI industry rejects the premise that market dilution of all works, or even a certain type of work, is a valid consideration under copyright law. This argument, albeit self-interested, has some merit under traditional fair use analysis. Fair use factor four, which considers whether a specific use potentially threatens the market value of the work(s) in suit may be narrowly construed to reject the kind of generalized market harm implied by GAI.

But as the quote above reveals, Judge Chhabria in Kadrey et al. v. Meta (not even one of the strongest cases against AI developers) recognizes the novelty of this technology to undermine the foundational purpose of copyright law.  He also states, “…by training generative AI models with copyrighted works, companies are creating something that often will dramatically undermine the market for those works, and thus dramatically undermine the incentive for human beings to create things the old-fashioned way.”[1] This recognition of far-reaching harm to the “incentive” foundation for copyright addresses an even broader question than the term “market dilution” implies.

A Broader Fourth Factor Analysis

In the Copyright Office’s report, the section called Market Dilution offers guidance for a reading of fair use factor four that is broad enough to address the fact that GAI outputs can harm the overall market for the same kind of works used in training. Here, I would endorse a view that broadens the fourth factor consideration, which traditionally only looks to potential harm to the copyright owner’s exclusive right to exploit the works in suit.

As argued in other posts, and in my comments to the USCO, the courts should place considerable weight in deciding whether the use at issue furthers the purpose of copyright. My submitted comment the Office chose to highlight states: “[G]enerative AI—if it does not produce market substitutes—primarily represents potential harm to authors and future authorship. . . .[T]he consideration in the context of ‘training’ should be expansive and doctrinal—namely that a potential threat to ‘authorship’ cannot, by definition, ‘promote the progress’ of ‘authorship.’”

I believe that dichotomy, novel to GAI, is precisely what the courts must resolve in order to prevent the technology from swallowing copyright law itself—perhaps especially where a given AI product does not output unlawful copies of works used in training.  The one consideration that rescues GAI products as promoting the purpose of copyright is where they are demonstrably “tools” for creators, but this goes to my overarching argument that the courts likely cannot obtain sufficient facts to discover whether the “tool” is constructive, destructive, or agnostic with regard to copyright’s purpose.

An AI tool used for ideation, for example, may further the purpose of copyright by helping the creator discover a new path from idea to protectable expression, but it is impossible for the court to assume this is the general purpose of the “tool.” The same product might just as easily be used in ways that are destructive to authorship.

For example, the vast majority of material produced and distributed will not be copyrightable due to the human authorship requirement for copyright rights to attach. Additionally, we are already seeing a large volume of AI “slop” distributed on platforms like Amazon and Spotify, and it is well established that driving traffic to garbage content is a profitable model for those willing to engage in the practice. Although a specific bucket of AI “slop,” when considered in a traditional fourth factor analysis, may not directly compete with any specific works of authorship, the courts should continue to give weight to the undeniable fact that a market flooded with “slop” does not in any sense promote copyright’s purpose and is most likely destructive to that purpose.

This view does not ignore or dismiss the creative and cultural potential of GAI as a means of expression. Many popular videos online are made with (presumably) human-authored scripts combined with AI generated AV material. That the expressions in these works will generally be unprotectable is a valid basis on which to find that the purpose of the AI product does not promote copyright’s purpose. But further, the fact that many of the creators of these works are not incentivized by copyright rights—they are motivated by the opportunity to share ad revenue with the platforms—means that these works, regardless of their qualitative value, live outside the copyright system. As such, works incentivized and enabled by a model other than copyright cannot reasonably be held to further the purpose of copyright.

In my view, these considerations look beyond the typical factor four analysis, and even beyond the ordinary concept of market dilution, to ask a fundamental question:  Can a technology built by mass copyright infringement properly make fair use of works when the product’s ultimate purpose is either destructive or irrelevant to the purpose of copyright law? I don’t think so.

Is Denial of Licensing for AI Training a Market Harm?

A recent post by Copyright Alliance CEO Keith Kupferschmid states that both Judge Alsup in Bartz and Judge Chhabria in Kadrey erred by too hastily concluding that authors are not entitled to license fees for the use of their works in AI training. On that assumption, both judges held that under factor four, the claimants could not show market harm due to the defendants’ failure to license. Kupferschmid writes:

Both judges are incorrect because they ignore the important realities that a robust emerging market for licensing of AI training material already exists. Licensing markets under the fourth factor may only be circular and non-cognizable when the market being considered is a potential licensing market and the judge is trying to determine whether that potential market is too speculative. But when there is an actual market that already exists, the circularity argument has no place and both judges were incorrect to summarily claim the argument is circular. 

Notably, Judge Chhabria, in rejecting the existence of a licensing market for AI training, cites Tresona Multimedia v. Burbank High School, but in addition to Kupferschmid’s point that a licensing market already exists for AI training, I am not sure the court’s reference to Tresona even applies. Judge Chhabria quotes from the opinion thus: “In every fair use case, the ‘plaintiff suffers a loss of a potential market if that potential [market] is defined as the theoretical market for licensing’ the use at issue in the case.” However, the next part of the opinion reads as follows:

…a copyright holder cannot prevent others from entering fair use markets merely ‘by developing or licensing a market for parody, news reporting, educational, or other transformative uses of its own creative work.’ (citation omitted)

This appears to tie the question of whether a licensing market is merely “theoretical” to a finding of whether the purpose of the use is indeed transformative. And although both the Kadrey and Bartz courts found those uses to be transformative, I believe those holdings are so tautological (i.e., lacking proper analysis) as to be ripe for significant challenge. Notably, at issue in Tresona was an educational use of small amounts of musical works—a paradigmatic fair use consideration, and one that may be as far from the implications of generative AI as we might imagine. “Further, the Warhol decision calls into question whether fair use cases like Tresona are still good law,” Kupferschmidt said to me by email.

The interplay between factors one and four, while inherent to the fair use analysis, reveals a vexing circularity in the context of GAI where the court is persuaded to find that the remarkable nature of the technology is transformative solely because the use appears to serve a “different purpose” than the works used. In addition to not fully aligning with Warhol, Judge Chhabria’s well-founded instincts about authored works “competing” with voluminous GAI works under factor four cannot be comfortably harmonized with the finding that the AI product serves a different purpose under factor one.  Clearly, if the purpose of the input material is to entertain and inform and the purpose of the “competing” output material is to entertain and inform, these are not different purposes.

The important difference, then, is that the input works are human authored, about which copyright law speaks volumes, while the output works are machine made, about which copyright law says almost nothing. In general, GAI no more adds to the productivity of copyright than the sea steadily eroding stone into an aesthetically pleasing “natural sculpture.” The courts need not attempt to foresee whether GAI will be socially beneficial or harmful but only find that in context to copyright law there are far more reasons to disfavor fair use than to favor it.

[1] I would have preferred that Judge Chhabria had not used “old fashioned,” which may be improperly read to mean “outdated” in contrast to AI generated works.