Chabon v. Chatbot:  About those ‘Shadow Libraries’

As many readers already know, another class-action lawsuit was filed on September 8 against OpenAI by book authors Michael Chabon, David Henry Hwang, Matthew Klam, Rachel Louise Snyder, and Ayelet Waldman on behalf of all authors similarly situated. The allegations are almost identical to the complaints in other class-action suits against various AI companies. I won’t repeat what I have already written about each allegation, but once again, I predict that if the court does not find unlawful reproduction in transient copies necessarily made in RAM, Open AI will likely prevail. Once again, this complaint alleges that the GPT model itself is an unlicensed “derivative work” of the entire corpus of books fed into it, but this does not seem to be a well-founded implication of the derivative works right under copyright law.

But one aspect of this complaint (as well as Tremblay et al.) is that Open AI is alleged to have obtained part of its database from known pirate repositories. In reference to one of the datasets used to train Chat GPT, the Chabon complaint states, “the only ‘internet-based books corpora’ that have ever offered that much material are infamous ‘shadow library’ websites, like Library Genesis (“LibGen”), Z-Library, Sci-Hub, and Bibliotik, which host massive collections of pirated books, research papers, and other text-based materials. The materials aggregated by these websites have also been available in bulk through torrent systems.” So, is the act of exploiting illegally obtained materials in this manner a violation of law?

Certainly, the Copyright Act does not address the issue. There is language about “lawfully made” copies in the context of first sale doctrine and certain exceptions for libraries. The only two uses of the words “lawfully obtained” in Title 17 pertain to acquisition of a computer program and permissible circumvention of technical protections for research purposes. So, nothing in the Copyright Act makes Open AI’s scraping “shadow libraries” an infringing act on its own, and there is no language in §107 on fair use that refers to lawfully making or obtaining material(s). This would be anathema since a fair use defense implies an unlicensed use.

Still, it seems wrong (probably because it is) to profit by exploiting another party’s unlawful possession of valuable materials. Under the criminal code (Title 18 §2315), it is a “federal offense to receive, possess, barter, sell, or dispose of stolen property with an aggregate value of $5,000 or more if the property crosses state lines.” The statute refers to physical property and not to exploiting databases full of pirated material. But if an AI developer knowingly exploits repositories replete with unlicensed copies of works, doesn’t that sound like it should be illegal?

This discussion reminds me a little bit of the rationale for the Protecting Lawful Streaming Act of 2020, which elevated the unauthorized public performance of works via streaming from a misdemeanor to a felony. After years of debate—and allegations by anti-copyright groups that felony streaming would be disastrous—Congress recognized that unlawful streaming is effectively a digital-age version of mass bootlegging physical copies, which had long been a felony. In fact, streaming is worse because it can reach a much larger black-market than any bootlegger distributing physical products ever could.

So, under a similar rationale by which Congress recognized that streaming digital repositories of unlicensed works is a felony, perhaps lawmakers might broaden the intent of Title 18 §2315 to prohibit mass exploitation of digital warehouses full of illegal copies of copyrighted works. Certainly, these warehouses contain materials with aggregate values in the tens of millions of dollars. Hence, any party that knowingly exploits these warehouses for financial gain might reasonably be liable under the criminal code.

Authors and artists are justifiably angry that their works are being used without permission to train generative AIs. And the fact that Chat GPT was allegedly trained in part with corpora of literary material acquired and stored by media pirates is salt in the wound to say the least. I don’t know what, if any, legal remedies might be proposed, but I am confident that it is generally wrong to profit from the intentional use of ill-gotten goods.


Photo by: onephoto

Generative AI Goes to the Opera

I think music is the purest artform because it is uniquely capable of provoking strong emotional responses without necessarily conveying meaning or information. Yes, one could say the same thing about abstract visual art, but I think the brain is hardwired to at least try to read meaning in visual expression and that this is not so with instrumental music. Moreover, I don’t think any medium is so universally provocative of human emotion as music.

It is admittedly cliché to talk about operatic arias provoking tears, but in my experience, they really do. In fact, one of my favorite arias is about a tear, aptly entitled “Una Furtiva Lagrima” (One Furtive Tear) from Gaetano Donizetti’s 1832 opera L’elisir Di Amore (The Elixir of Love). I do have a personal relationship with this song because it was first introduced to me by my late father-in-law, a tenor who sang with several U.S. opera companies, served as artist in residence at The Israeli Nation Opera, and sang for Pope John Paul II in 1988. I wish had a digital version of his “Una Furtiva Lagrima” to share because it is, in classical terms, the shiznit.

But I was thinking about that aria for this post because, notwithstanding the familial connection, nothing external to the music influences its effect on me. I am not an expert on opera or Donizetti, and I do not fully understand the Italian libretto. Hence, the mechanics by which the score and the tenor’s performance reach through this curmudgeon’s crusty exterior to trigger an emotional response can be boiled down to a science, which means that a similar experience can be created by a generative AI. And so, the elephant in the room asks the obvious question:  Will the provenance of a work matter to the people who experience it?

I recognize that music by generative AI is already responding to this question, but these early sprouts in the market do not tell us what the broader cultural effects might be in a future without Donizettis, Domingos, or orchestras. One valid prediction could be that it won’t matter to the audience experiencing the music whether it was generated by a machine or another human. If a song produces spontaneous tears or laughter or a desire to dance, then who cares if it was made in a lab rather than by charming Liverpudlians sweating it out in a London studio?

Most Artists Are Not Performers

This conversation requires that we make a distinction between performance and composition. In other posts, when I’ve scorned the idea of machines replacing artists, I have generally referred to performance and drawn analogies to sports. One that seems to resonate in conversation is my NASCAR example because this is basically watching machines move in circles and waiting to see which machine finishes the requisite number of circles first. This lifeless description makes the point that without the people in the drivers’ seats and pit crews—humans who are largely hidden from view during the race—NASCAR would be about as interesting as watching an oil pump bob its mechanical head at the ground.

I believe our desire, or need, to experience performance—whether it’s Blake Morgan playing his music or Coco Gauff winning the Women’s US Open—mitigates AI’s power to usurp the role of many artists. But if this is true, the rule only applies when composition and performance are deeply intertwined, as with singer/songwriters like the recently late Jimmy Buffett. An AI “Caribbean-Drunk-Rock-n-Roll-Music”[1] generator could never foster the whole experience that became the Buffett brand. But could this ersatz “Margaritaville” mixer compose the equivalent of a new “Come Monday,” and if so, would it matter to future listeners who have no idea what the AI “learned” from Jimmy?

Most creators are “composers” and not “performers,” often as removed from the audience experiencing their work as I am from Donizetti while listening to his aria in 2023. And frankly, Donizetti, who died in 1848, is hardly more obscure to the average listener than Rod Temperton, who died in 2016 after writing some of the most popular songs of the 1970s and 80s including several of Michael Jackon’s biggest hits. Never in my teen years was I aware of Mr. Temperton’s role in all those songs.

So, keeping the focus on the composers, authors, painters, photographers, filmmakers et al. who do not perform, is there some anthropological reason to believe (hope) that artists will not be replaced by machines making music, books, visual arts, etc.? I understand that there are practical reasons why AIs may not get there at scale, but the question I’m asking is more about us than about the technology. Will the science that makes music provocative continue to work on the human listener, if future compositions are produced by things that cannot feel heartache or longing or humor, etc.? Put differently, will the novelty of generative AI wear off because the compositions it produces will become flat, bloodless, and disposable?

In my book, I wondered why an advanced AI (one that can make even semi-autonomous decisions) would bother to produce “art” upon reaching a certain threshold in its so-called intelligence. If humans make art because it’s one way we confront, synthesize, and respond to the human experience, then perhaps the “smarter” the AI becomes, the more likely it is to realize that it has nothing to say because it has no experience. Or does the robot begin to create works in response to the robot experience and ignore its instructions to produce songs or novels or pictures for human consumption? I doubt it, but if this does happen, we can be sure that some humans will form a cult to follow the new bot prophet.

But I’m not really answering the thesis question, am I? Because I have no idea. I want to believe that the question was answered by Ian Malcom (Jeff Goldblum) in Jurassic Park when he warned that nature finds a way.* Only instead of dinosaur nature triumphing over laboratory safeguards to keep them contained, it would be human nature instinctively rejecting synthetic “art” for reasons that are likewise ineffable. For better or worse the AI experiment, like Jurassic Park, has begun, and we’ll have to wait and see who gets eaten. So, perhaps the new version of the Turing Test should not be whether the computer can make you believe it’s human, but whether it can provoke a furtive tear and then ask whether you mind that it is not human.


[1] Buffett’s own description from his live album You Had to Be There.

*Thanks to comment by Bob Hill. Malcolm says “Life finds a way.” I edited the text to retain the point but drop the quotation marks.

Photo in collage: Thomas O’Leary in The Tales of Hoffmann.

The Mugshot Heard ‘round the World

It was no surprise that the mugshot was immediately copied onto tees, hats, coffee mugs, etc. and sold to Americans who see either a martyr or a traitor in the same image. It was also no surprise that Team Trump produced merch of its own to sell for campaign (a.k.a. criminal defense) fundraising purposes. But these and other uses of the photograph have fostered some legal discussions on chat boards and elsewhere as to who, if anyone, has the right to control the exploitation of the mugshot. And so, I offer my own takes for what they’re worth.

Who Owns the Copyrights in the Image?

This is actually two questions:  1) is the Trump mugshot copyrightable at all? and 2) if so, who would be the owner of the copyright? Opinions will vary, but in my view, there are several factors that militate against enforceable copyright in this photograph, which is tantamount to having no copyright at all. If any party could own the copyright, it would logically be the State of Georgia or Fulton County, but aside from the fact that neither entity is likely to file a registration application for the photo with the Copyright Office, there is arguably no basis for finding sufficient originality in the image.

The mugshot photo station at the jailhouse is presumably as static as a surveillance camera—arranged to capture the same, fact-intensive photo for a highly utilitarian, informative purpose. No human (e.g., officer or clerk) can reasonably claim to have made any creative choices to produce original expression in the Trump mugshot, and this militates against copyright rights, which would then automatically transfer to the state or county employer. If there is any expression in the image at all, it is arguably Trump’s “creative” choice to make the angry face. But although I have explored the question of co-authorship by subjects in photographs, this is 1) a thought experiment outside the bounds of case law; and 2) a theory that would likely find less foundation in an image that is more factual than expressive in nature.

For these reasons alone, I believe the image would not be copyrightable, even if the state entity were to try to register the photograph with the Copyright Office. But no matter what, there is no legal authority under which Trump could own the copyright.

Can the Trump Campaign Control the Merch?

On August 29, Trump campaign adviser Chris LaCivita posted on X, “If you are a campaign, PAC, scammer and you try raising money off the mugshot of @realDonaldTrump and you have not received prior permission…WE ARE COMING AFTER YOU…you WILL NOT SCAM DONORS.”

Notwithstanding the tongue-biting irony of Team Trump using the word scam, LaCivita’s message could be read as a valid warning to any parties that might pretend to be the Trump campaign, but that would be an odd statement in regard to the mugshot because this type of fraud has nothing to do with use of the photograph per se. If, instead, LaCivita means to imply that the Trump campaign has an exclusive right to sell “official” mugshot merchandise for commercial purposes—or to prevent use of the image to raise funds in opposition to Trump—then he’s dead wrong on the law, as that crowd so often is.

Trademark Law Does Nothing for Trump

Although it is permissible to register trademarks in certain words or images used in political campaigns (e.g., slogans or logos), there are both administrative and doctrinal reasons why the Trump campaign could not claim the mugshot as a service mark. As a practical matter, the trademark claimant must use the relevant mark in trade when applying for protection and then go through a rather lengthy process to affirm the mark remains in use—and use in a specific class (or classes) of goods and/or services.

But in this case, the instant the mugshot was shared with the world, it conveyed irreconcilably divergent meanings to the public. So, under trademark practice, could Trump assert the exclusive right to use the “mark” in a class called Multiply Indicted, Seditious Former Presidents? Probably not since no such class exists. But that’s generally what the image conveys to millions of Americans, and the purpose of trademark is to protect the earned integrity of brands, not to burnish the reputations of politicians reviled by more than half the population.

What About Trump’s Likeness?

It may not be Trump’s mugshot as IP, but it is certainly his mug, and doesn’t his right of publicity (ROP) allow him to control how his likeness is used? As discussed in context to artificial intelligence, ROP laws are statutory in half the states, common law elsewhere, and there is no federal ROP statute. Most importantly, though, ROP generally applies to commercial use of an individual’s likeness for endorsement or advertising purposes. Thus, Susan Scafidi, founder of the Fashion Law Institute is off the mark, as quoted in the New York Times stating, “Trump could, in theory, attempt to shut down sales of merch with his mug shot, not unlike the way Obama objected to appearing on a Weatherproof Garment Company billboard…”

I believe this is incorrect. Unauthorized use of a likeness (even of a political figure) for commercial advertising is likely to be a paradigmatic violation of ROP. So, if an entity were to use the Trump mugshot to promote its goods or services, Trump should have a strong legal foundation for stopping that use.[1] By contrast, reproducing the mugshot for the purpose of mocking, criticizing, or downright hating any political figure is protected speech at the core of the First Amendment, and Trump would have no legal foundation to enjoin such uses.

But what if the mugshot is reproduced (on merch or elsewhere) without accompanying commentary? If I walk through town wearing a tee shirt with the unaltered mugshot on it, observers who don’t know me would have no idea whether I am celebrating or denouncing the Georgia arraignment. So, does this ambiguity alter the First Amendment consideration such that Trump would have any grounds to stop the production of merchandise that merely reproduces the photo? Again, I would say no if only because the mugshot is a factual statement of extraordinary newsworthy value to the public. Thus, the production and distribution of merchandise bearing no communication other than the image should still be protected by the speech and press rights, even if the right of redress is not implicated.

So, that’s my 50 cents on some of the legal discussion surrounding this image, which may one day be more widely reproduced than Alberto Korda’s photograph of Che Guevara.[2] Of course, this is all nerdy food for thought because it’s hard to imagine that any of these questions will ever be presented in court. Even if Team Trump could show standing, they have bigger sheep to fleece and zero hope of controlling the perception of millions that a mugshot is usually just a photograph of a criminal.


[1] It is of course possible to blur the line between a company’s politics and its marketing, which would result in a fact-intensive inquiry into the matter. Likewise, a not-for-profit could promote a policy message that Trump does not endorse and use the mugshot to illustrate the opposition, and this should not be a violation of ROP.

[2] Ironically, this is actual and rampant infringement of the photographer’s copyright rights.