Tag: fair use

Court Rules AI Training is Fair Use

Court rules that using copyrighted works to train AI is fair use. Kadrey et al. v. Meta Platforms.

Just days after the first major fair use ruling in a generative-AI case, a second court has determined that using copyrighted works to train AI is fair use. Kadrey et al. v. Meta Platforms, No. 3:23-cv-03417-VC (N.D. Cal. June 25, 2025).

The Kadrey v. Meta Platforms Lawsuit

I previously wrote about this lawsuit here and here.

Meta Platforms owns and operates social media services including Facebook, Instagram, and WhatsApp. It is also the developer of a large language model (LLM) called “Llama.” One of its releases, Meta AI, is an AI chatbot that utilizes Llama.

To train its AI, Meta obtained data from a wide variety of sources. The company initially pursued licensing deals with book publishers. It turned out, though, that in many cases, individual authors owned the copyrights. Unlike music, no organization handles collective licensing of rights in book content. Meta then downloaded shadow library databases. Instead of licensing works in the databases, Meta decided to just go ahead and use them without securing licenses. To download them more quickly, Meta torrented them using BitTorrent.

Meta trained its AI models to prevent them from “memorizing” and outputting text from the training data, with the result that no more than 50 words and punctuation marks from any given work were reproduced in any given output.

The plaintiffs named in the Complaint are thirteen book authors who have published novels, plays, short stories, memoirs, essays, and nonfiction books. Sarah Silverman, author of The Bedwetter; Junot Diaz, author of The Brief Wondrous Life of Oscar Wao; and Andrew Sean Greer, author of Less, are among the authors named as plaintiffs in the lawsuit. The complaint alleges that Meta downloaded 666 copies of their books without permission and states claims for direct copyright infringement, vicarious copyright infringement, removal of copyright management information in violation of the Digital Millennium Copyright Act (DMCA), and various state law claims. All claims except the ones for direct copyright infringement and violation of the DMCA were dismissed in prior proceedings.

Both sides moved for summary judgment on fair use with respect to the claim that Meta’s use of the copyrighted works to train its AI infringed copyrights. Meta moved for summary judgment on the DMCA claims. Neither side moved for summary judgment on a claim that Meta infringed copyrights by distributing their works (via leeching or seeding).

On June 25, 2025 Judge Chhabria granted Meta’s motion for summary judgment on fair use with respect to AI training; reserved the motion for summary judgment on the DMCA claims for decision in a separate order, and held that the claim of infringing distribution via leeching or seeding “will remain a live issue in the case.”

Judge Chhabria’s Fair Use Analysis

Judge Chhabria analyzed each of the four fair use factors. As is the custom, he treated the first (Character or purpose of the use) and fourth (Effect on the market for the work) factors as the most important of the four.

He disposed of the first factor fairly easily, as Judge Alsup did in Bartz v. Anthropic, finding that the use of copyrighted works to train AI is a transformative use. This finding weighs heavily in favor of fair use. The purpose of Meta’s AI tools is not to generate books for people to read. Indeed, in this case, Meta had installed guardrails to prevent the tools from generating duplicates or near-duplicates of the books on which the AI was trained. Moreover, even if it could allow a user to prompt the creation of a book “in the style of” a specified author, there was no evidence that it could produce an identical work or a work that was substantially similar to one on which it had been trained. And writing styles are not copyrightable.

Significantly, the judge held that the use of shadow libraries to obtain unauthorized copies of books does not necessarily destroy a fair use defense. When the ultimate use to be made of a work is transformative, the downloading of books to further that use is also transformative, the judge wrote. This ruling contrasts with other judges who have intimated that using pirated copies of works weighs against, or may even prevent, a finding of fair use.

Unlike some judges, who tend to consider the fair use analysis over and done if transformative use is found, Judge Chhabria recognized that even if the purpose of the use is transformative, its effect on the market for the infringed work still has to be considered.

3 Ways of Proving Adverse Market Effect

The Order lays out three potential kinds of arguments that may be advanced to establish the adverse effect of an infringing use on the market for the work:

  1. The infringing work creates a market substitute for the work;
  2. Use of the work to train AI without permission deprives copyright owners of a market for licenses to use their works in AI training;
  3. Dilution of the market with competing works.

Market Substitution

In this case, direct market substitution could not be established because Meta had installed guardrails that prevented users from generating copies of works that had been used in the training. Its AI tools were incapable of generating copies of the work that could serve as substitutes for the authors’ works.

The Market for AI Licenses

The court refused to recognize the loss of potential profits from licensing the use of a work for AI training purposes as a cognizable harm.

Market Dilution

The argument here would be that the generation of many works that compete in the same market as the original work on which the AI was trained dilutes the market for the original work. Judge Chhabria described this as indirect market substitution.

The copyright owners in this case, however, focused on the first two arguments. They did not present evidence that Meta’a AI tools were capable of generating books; that they do, in fact, generate books; or that the books they generate or are capable of generating compete with books these authors wrote. There was no evidence of diminished sales of their books.

Market harm cannot be assumed when generated copies are not copies that can serve as substitutes for the specific books claimed to have been infringed. When the output is transformative, as it was in this case, market substitution is not self-evident.

Judge Chhabria chided the plaintiffs for making only a “half-hearted argument” of a significant threat of market harm. He wrote that they presented “no meaningful evidence on market dilution at all.”

Consequently, he ruled that the fourth fair use factor favored Meta.

Conclusion

The decision in this case is as significant for what the court didn’t do as it is for what it did. It handed a fair use victory to Meta. At the same time, though, it did not rule out a finding that training AI tools on copyrighted works is not fair use in an appropriate case. The court left open the possibility that a copyright owner might prevail on a claim that training AI on copyrighted works is not fair use in a different case. And it pointed the way, albeit in dictum, namely, by making a strong showing of market dilution.

That claim is not far-fetched. https://www.wired.com/story/scammy-ai-generated-books-flooding-amazon/

AI OK; Piracy Not: Bartz v. Anthropic

Anthropic also acquired infringing copies of works from pirate sites. Judge Alsup ruled that these, and uses made from them, are not fair use.

A federal judge has issued a landmark fair use decision in a generative-AI copyright infringement lawsuit.

In a previous blog post, I wrote about the fair use decision in Thomson Reuters v. ROSS. As I explained there, that case involved a search-and-retrieval AI system, so the holding was not determinative of fair use in the context of generative AI. Now we finally have a decision that addresses fair use in the generative-AI context.

Bartz et al. v. Anthropic PBC

Anthropic is an AI software firm founded by former OpenAI employees. It offers a generative-AI tool called Claude. Like other generative-AI tools, Claude mimics human conversational skills. When a user enters a text prompt, Claude will generate a response that is very much like one a human being might make (except it is sometimes more knowledgeable.) It is able to do this by using large language models (LLMs) that have been trained on millions of books and texts.

Adrea Bartz, Charles Graeber, and Kirk Wallace Johnson are book authors. In August 2024, they sued Anthropic, claiming the company infringed the copyrights in their works. Specifically, they alleged that Anthropic copied their works from pirated and purchased sources, digitized print versions, assembled them into a central library, and used the library to train LLMs, all without permission. Anthropic asserted, among other things, a fair use defense.

Earlier this year, Anthropic filed a motion for summary judgment on the question of fair use.

On June 23, 2025, Judge Alsup issued an Order granting summary judgment in part and denying it in part. It is the first major ruling on fair use in the dozens of generative-AI copyright infringement lawsuits that are currently pending in federal courts.

The Order includes several key rulings.

Books

Digitization

Anthropic acquired both pirated and lawfully purchased printed copies of copyright-protected works and digitized them to create a central e-library. Authors claimed that making digital copies of their works infringed the exclusive right of copyright owners to reproduce their works. (See 17 U.S.C. 106.)

In the process of scanning print books to create digital versions of them, the print copies were destroyed. Book bindings were stripped so that each individual page could be scanned. The print copies were then discarded. The digital copies were not distributed to others. Under these circumstances, the court ruled that making digital versions of print books is fair use.

The court likened format to a frame around a work, as distinguished from the work itself. As such, a digital version is not a new derivative work. Rather, it is a transformative use of an existing work. So long as the digital version is merely a substitute for a print version a person has lawfully acquired, and so long as the print version is destroyed and the digital version is not further copied or distributed to others, then digitizing a printed work is fair use. This is consistent with the first sale doctrine (17 U.S.C. 109(a)), which gives the purchaser of a copy of a work a right to dispose of that particular copy as the purchaser sees fit.

In short, the mere conversion of a lawfully acquired print book to a digital file to save space and enable searchability is transformative, and so long as the print version is destroyed and the digital version is not further copied or distributed, it is fair use.

AI Training Is Transformative Fair Use

The authors did not contend that Claude generated infringing output. Instead, they argued that copies of their works were used as inputs to train the AI. The Copyright Act, however, does not prohibit or restrict the reading or analysis of copyrighted works. So long as a copy is lawfully purchased, the owner of the purchased copy can read it and think about it as often as he or she wishes.

[I]f someone were to read all the modern-day classics because of their exceptional expression, memorize them, and then emulate a blend of their best writing, would that violate the Copyright Act? Of course not.

Order.

Judge Alsup described AI training as “spectacularly” transformative.” Id. After considering all four fair use factors, he concluded that training AI on lawfully acquired copyright-protected works (as distinguished from the initial acquisition of copies) is fair use.

Pirating Is Not Fair Use

In addition to lawfully purchasing copies of some works, Anthropic also acquired infringing copies of works from pirate sites. Judge Alsup ruled that these, and uses made from them, are not fair use. The case will now proceed to trial on the issue of damages resulting from the infringement.

Conclusion

Each of these rulings seems, well, sort of obvious. It is nice to have the explanations laid out so clearly in one place, though.

 

Photographers’ Rights

The Second Circuit Court of Appeals reversed a trial judge’s dismissal of a photographer’s copyright infringement complaint, holding that because “fair use” was not clearly established on the face of the complaint, the district court should not have dismissed the complaint sua sponte. Romanova v. Amilus, Inc.

Romanova v. Amilus, Inc., No. 23-828 (2nd Cir., May 23, 2025)

The Second Circuit Court of Appeals reversed a trial judge’s dismissal of a photographer’s copyright infringement complaint, holding that because “fair use” was not clearly established on the face of the complaint, the district court should not have dismissed the complaint sua sponte.

European grass snake photograph illustrating copyright infringement article by Cokato Minnesota attorney Thomas James

Photographer Jana Romanova created a photograph of a woman with a snake wrapped around her left hand and another snake crawling up her torso. (Not the one pictured here.) She licensed it to National Geographic Magazine for a single use. According to the complaint, Amilus, Inc. allegedly made a copy of the photograph and published it to its website. Romanova allegedly sent notifications demanding the removal of the photograph from the website. The defendant allegedly did not respond. This lawsuit followed.

The defendant allegedly did not appear or respond to the complaint, so Romanova moved for the entry of default judgment. Rather than grant a default judgment, however, the district court judge sua sponte ordered Romanova to show cause why the court should not dismiss the case on the grounds that the defendant’s use of the photograph was fair use. Although fair use is an affirmative defense, which defendants have the burden of asserting and proving, the judge opined that the fair use defense did not need to be pleaded because the judge believed the fair use defense was “clearly established on the fact of the complaint.“

Romanova appealed. The Second Circuit Court of Appeals reversed, effectively allowing the infringement claim to go forward.

Fair Use

In its decision, the Second Circuit Court of Appeals clarified how courts are to interpret and apply the four-factor “fair use” test outlined in the Copyright Act, 17 U.S.C. § 107 (purpose and character of the use; nature of the work; amount and substantiality of the portion copied; and the effect on the market for the work.)

The district court concluded that the defendant’s publication of the photograph communicated a different message than what the photographer intended. According to the district court, the purpose of the publication in the National Geographic was “to showcase persons in [her] home country of Russia that kept snakes as pets, specifically to capture pet snakes in common environments that are more associated with mainstream domesticated animals.” The district court found that the purpose of the defendant’s publication was to communicate a message about “the ever-increasing amount of pet photography circulating online.“

Apparently the district court was under the impression that the use of a copyright-protected work for any different purpose, or to communicate any different message, is “transformative” and therefore “fair use.” The Court of Appeals clarified that is not the case. In addition to alleging and proving the use was for a different purpose or conveyed a different meaning, a defendant seeking to establish a fair use defense must also allege and prove a justification for the copying.

Examples of purposes that may justify copying a work include commentary or criticism of the copied work, or providing information to the public about the copied work, in circumstances where the copy does not become a substitute for the work. (See, e.g., Authors Guild v. Google, Inc., 804 F.3d 202, 212 (2d Cir. 2015).) Copying for evidentiary purposes (such as to support a claim that the creator of the work published a defamatory statement) can also be a valid justification to support a fair use defense. Creating small, low-resolution copies of images (“thumbnails”) may be justified when the purpose is to facilitate Internet searching. (Perfect 10 v. Amazon.com, 508 F.3d 1146, 1165 (9th Cir. 2007). Facilitating blind people’s access to a work may provide a justification for converting it into a format that blind people can read. (Authors Guild v. HathiTrust, 755 F.3d 87, 97 (2d Cir. 2014).

The Court cited other examples of potential justifications for copying. The Court admonished, however, that the question whether justification exists is a fact-specific determination that must be made on a case-by-case basis.

[J]ustification is often found when the copying serves to critique, or otherwise comment on, the original, or its author, but can also be found in other circumstances, such as when the copying provides useful information about the original, or on other subjects, usually in circumstances where the copying does not make the expressive content of the original available to the public.

Romanova, supra.

The only “justification” the district court cited for the copying was that it believed the defendant merely wanted to illustrate its perception of a growing trend to publish photographs of people with pets. “Little could remain of an author’s copyright protection if others could secure the right to copy and distribute a work simply by asserting some fact about the copied work,” the Court observed. The defendant’s publication of the copy did not communicate criticism or commentary on the original photograph or its author, or any other subject, the Court held.

The Court held that the remaining three fair use factors also militated against a finding of fair use.

Sua Sponte Dismissal for “Fair Use”

Justice Sullivan filed a concurring opinion. He would have reversed on procedural grounds without reaching the substantive issue. Specifically, Justice Sullivan objected to the trial judge’s raising of the fair use defense sua sponte on behalf of a non-appearing defendant. Normally, if a complaint establishes a prima case for relief, the court does not consider affirmative defenses (such as fair use) unless the defendant asserts them. That is to say, fair use is an affirmative defense; the defendant, not the plaintiff, bears the burden of proof.

Conclusion

Appeals courts continue to rein in overly expansive applications of “transformative” fair use by the lower courts. Here, the Court of Appeals soundly reasoned that merely being able to articulate an additional purpose served by publishing an author’s entire work, unchanged, will not, by itself, suffice to establish either transformative use or fair use.

 

Fair Use Decision in Thomson Reuters v. Ross

A court has handed down the first known ruling (to me, anyway) on “fair use” in the wave of copyright infringement lawsuits against AI companies that are pending in federal courts.

A court has handed down the first known ruling (to me, anyway) on “fair use” in the wave of copyright infringement lawsuits against AI companies that are pending in federal courts. The ruling came in Thomas Reuters v. Ross. Thomas Reuters filed this lawsuit against Ross Intelligence back in 2020, alleging that Ross trained its AI models on Westlaw headnotes to build a competing legal research tool, infringing numerous copyrights in the process. Ross asserted a fair use defense.

In 2023, Thomson Reuters sought summary judgment against Ross on the fair use defense. At that time, Judge Bibas denied the motion. This week, however, the judge reversed himself, knocking out at least a major portion of the fair use defense.

Ross had argued that Westlaw headnotes are not sufficiently original to warrant copyright protection and that even if they are, the use made of them was “fair use.” After painstakingly reviewing the headnotes and comparing them with the database materials, he concluded that 2,243 headnotes were sufficiently original to receive copyright protection, that Ross infringed them, and that “fair use” was not a defense in this instance because the purpose of the use was commercial and it competed in the same market with Westlaw. Because of that, it was likely to have an adverse impact on the market for Westlaw.

While this might seem to spell the end for AI companies in the many other lawsuits where they are relying on a “fair use” defense, that is not necessarily so. As Judge Bibas noted, the Ross AI was non-generative. Generative AI tools may be distinguishable in the fair use analysis.

I will be presenting a program on Recent Developments in AI Law in New Jersey this summer. This one certainly will merit mention. Whether any more major developments will come to pass between now and then remains to be seen.

New AI Copyright Infringement Lawsuit

Another copyright and trademark infringement lawsuit against an AI company was filed this week. This one pits news article publishers Advance Local Media, Condé Nast, The Atlantic, Forbes Media, The Guardian, Business Insider, LA Times, McClatchy Media Company, Newsday, Plain Dealer Publishing Company, POLITICO, The Republican Company, Toronto Star Newspapers, and Vox Media against AI company Cohere.

The complaint alleges that Cohere made unauthorized use of publisher content in developing and operating its generative AI systems, infringing numerous copyrights and trademarks. The plaintiffs are seeking an injunction and monetary damages.

More copyright stories here.

Top Copyright Cases of 2024

Many AI-related copyright lawsuits continued to proceed through the courts

Warner Chappell Music Inc. v. Nealy

The Copyright Act imposes a three-year period of limitations for copyright infringement claims. There has been a split in the circuits about whether this means that damages could be claimed only for infringement occurring during the three-year period or whether damages could be recovered for earlier acts of infringement so long as the claim is timely filed.

The issue arises in cases where a claimant invokes the discovery rule. The general rule is that a limitations period runs from the date of the act giving rise to the cause of action. The discovery rule, by contrast, measures the limitations period from the date the infringing act is discovered. Thus, for example, if an infringing act occurred in 2012 but the copyright owner did not learn about it until 2022, then under the traditional rule, the claim would be time-barred. Under the discovery rule, it would not be.

The Court’s holding means that if the discovery rule applies in the jurisdiction where suit is filed, and a claimant properly invokes it, then damages are not limited to the three years preceding suit. Rather, any damages incurred since the date of the infringint act are recoverable.

The Court did not rule on the validity of the discovery rule.

Warner Chappell Music Inc.. v. Nealy, 601 U.S. ____ (2024). Read more here.

 Hachette Book Group Inc. v. Internet Archive

I wrote about this case back in 2022, when it was at the summary judgment stage in the district court for the Southern District of New York. The complaint, filed by book publishers, alleged that the Internet Archive made digital copies of over a million print books and then freely distributed the copies to members of the public, all without the permission of the copyright owners. In 2023, the district judge ruled in favor of the publishers, holding that the enterprise was not “fair use.” This year, the Second Circuit Court of Appeals affirmed the decision.

To some, the decision might seem like a no-brainer. Copying other people’s books and giving them away for free, without the copyright owners’ permission, sounds like core copyright infringement, right? Yet, before the Warhol v. Goldsmith decision in 2023, courts had been applying such an expansive view of the “transformative use” branch of fair use that some people thought that making digital copies of a print book was categorically “transformative” and therefore fair use. This decision makes it clear that no, it isn’t.

The Internet Archive has said it will not appeal the decision to the United States Supreme Court.

Hachette Book Group Inc. et al. v. Internet Archive, No. 23-1260 (2nd Cir. 2024)

Griner v. King

U.S. Representative Steve King’s campaign committee used a copyright-protected photograph in his campaign without permission. King’s committee had argued fair use and that it had an “implied license” to use the image because it had been widely circulated as a meme on the Internet. The Eighth Circuit Court of Appeals upheld an Iowa jury’s verdict for the copyright owner.

Griner et al. v. King et al., No. 23-2117, (8th Cir. 2024)

The Intercept Media v. OpenAI

This isn’t really a momentous decision, in terms of precedential value, but it is the first major victory for Big AI in the plethora of AI-related lawsuits they are facing.

The Intercept Media, Inc. sued OpenAI and Microsoft Corporation for alleged Digital Millennium Copyright Act (DMCA) violations in connection with training the AI tool, ChatGPT. The defendants filed a motion to dismiss. On November 21, 2024 the New York court dismissed claims against Microsoft with prejudice. The court dismissed the 17 U.S.C. § 1202(b)(3) claim against OpenAI but allowed the claim under 17 U.S.C. §1202(b)(1) to proceed.  

Section 1202(b)(1) prohibits unauthorized removal or alteration of copyright management information, including author information and the copyright notice.

The Intercept Media Inc. v. OpenAI Inc., No. 1:24-cv-01515, (S.D.N.Y. Nov. 21, 2024).

Stay tuned…

Many AI-related copyright lawsuits continued to proceed through the courts in 2024, with decisions expected in 2025 or later.

 

AI Lawsuits Roundup

A status update on 24 pending lawsuits against AI companies – what they’re about and what is happening in court – prepared by Minnesota copyright attorney Thomas James.

A very brief summary of where pending AI lawsuits stand as of February 28, 2024. Compiled by Minnesota attorney Thomas James.

Thomson Reuters v. Ross, (D. Del. 2020)

Filed May 6, 2020. Thomson Reuters, owner of Westlaw, claims that Ross Intelligence infringed copyrights in Westlaw headnotes by training AI on copies of them. The judge has granted, in part, and denied, in part, motions for summary judgment. The questions of fair use and whether the headnotes are sufficiently original to merit copyright protection will go to a jury to decide.

Thaler v. Perlmutter (D.D.C. 2022).

Complaint filed June 2, 2022. Thaler created an AI system called the Creativity Machine. He applied to register copyrights in the output he generated with it. The Copyright Office refused registration on the ground that AI output does not meet the “human authorship” requirement. He then sought judicial review. The district court granted summary judgment for the Copyright Office. In October, 2023, he filed an appeal to the District of Columbia Circuit Court of Appeals (Case no. 23-5233).

Doe v. GitHub, Microsoft, and OpenAI (N.D. Cal. 2022)

Complaint filed November 3, 2022. Software developers claim the defendants trained Codex and Copilot on code derived from theirs, which they published on GitHub. Some claims have been dismissed, but claims that GitHub and OpenAI violated the DMCA and breached open source licenses remain. Discovery is ongoing.

Andersen v. Stability AI (N.D. Cal. 2023)

Complaint filed January 13, 1023. Visual artists sued Midjourney, Stability AI and DeviantArt for copyright infringement for allegedly training their generative-AI models on images scraped from the Internet without copyright holders’ permission. Other claims included DMCA violations, publicity rights violations, unfair competition, breach of contract, and a claim that output images are infringing derivative works. On October 30, 2023, the court largely granted motions to dismiss, but granted leave to amend the complaint. Plaintiffs filed an amended complaint on November 29, 2023. Defendants have filed motions to dismiss the amended complaint. Hearing on the motion is set for May 8, 2024.

Getty Images v. StabilityAI (U.K. 2023)

Complaint filed January, 2023. Getty Images claims StabilityAI scraped images without its consent. Getty’s complaint has survived a motion to dismiss and the case appears to be heading to trial.

Getty Images v. Stability AI (D. Del.)

Complaint filed February 3, 2023. Getty Images alleges claims of copyright infringement, DMCA violation and trademark violations against Stability AI. The judge has dismissed without prejudice a motion to dismiss or transfer on jurisdictional grounds. The motion may be re-filed after the conclusion of jurisdictional discovery, which is ongoing.

Flora v. Prisma Labs (N.D. Cal.)

Complaint filed February 15, 2023. Plaintiffs allege violations of the Illinois Biometric Privacy Act in connection with Prisma Labs’ collection and retention of users’ selfies in AI training. The court has granted Prisma’s motion to compel arbitration.

Kyland Young v. NeoCortext (C.D. Cal. 2023)

Complaint filed April 3, 2023. This complaint alleges that AI tool Reface used a person’s image without consent, in violation of the person’s publicity rights under California law. The court has denied a motion to dismiss, ruling that publicity rights claims are not preempted by federal copyright law. The case has been stayed pending appeal.

Walters v. OpenAI (Gwinnett County Super. Ct. 2023), and Walters v. OpenAI (N.D. Ga. 2023)

Gwinnett County complaint filed June 5, 2023.

Federal district court complaint filed July 14, 2023.

Radio talk show host sued OpenAI for defamation. A reporter had used ChatGPT to get information about him. ChatGPT wrongly described him as a person who had been accused of fraud. In October, 2023, the federal court remanded the case to the Superior Court of Gwinnett County, Georgia.  On January 11, 2024, the Gwinnett County Superior Court denied OpenAI’s motion to dismiss.

P.M. v. OpenAI (N.D. Cal. 2023).

Complaint filed June 28, 2023. Users claim OpenAI violated the federal Electronic Communications Privacy Act and California wiretapping laws by collecting their data when they input content into ChatGPT. They also claim violations of the Computer Fraud and Abuse Act. Plaintiffs voluntarily dismissed the case on September 15, 2023. See now A.T. v. OpenAI (N.D. Cal. 2023) (below).

In re OpenAI ChatGPT Litigation (N.D. Cal. 2023)

Complaint filed June 28, 3023. Originally captioned Tremblay v. OpenAI. Book authors sued OpenAI for direct and vicarious copyright infringement, DMCA violations, unfair competition and negligence. Both input (training) and output (derivative works) claims are alleged, as well as state law claims of unfair competition, etc. Most state law and DMCA claims have been dismissed, but claims based on unauthorized copying during the AI training process remain. An amended complaint is likely to come in March. The court has directed the amended complaint to consolidate Tremblay v. OpenAI, Chabon v. OpenAI, and Silverman v. OpenAI.  

Battle v. Microsoft (D. Md. 2023)

Complaint filed July 7, 2023. Pro se defamation complaint against Microsoft alleging that Bing falsely described him as a member of the “Portland Seven,” a group of Americans who tried to join the Taliban after 9/11.

Kadrey v. Meta (N.D. Cal. 2023)

Complaint filed July 7, 2023. Sarah Silverman and other authors allege Meta infringed copyrights in their works by making copies of them while training Meta’s AI model; that the AI model is itself an infringing derivative work; and that outputs are infringing copies of their works. Plaintiffs also allege DMCA violations, unfair competition, unjust enrichment, and negligence. The court granted Meta’s motion to dismiss all claims except the claim that unauthorized copies were made during the AI training process. An amended complaint and answer have been filed.

J.L. v. Google (N.D. Cal. 2023)

Complaint filed July 11, 2023. An author filed a complaint against Google alleging misuse of content posted on social media and Google platforms to train Google’s AI Bard. (Gemini is the successor to Google’s Bard.) Claims include copyright infringement, DMCA violations, and others. J.L. filed an amended complaint and Google has filed a motion to dismiss it. A hearing is scheduled for May 16, 2024.

A.T. v. OpenAI (N.D. Cal. 2023)

Complaint filed September 5, 2023. ChatGPT users claim the company violated the federal Electronic Communications Privacy Act, the Computer Fraud and Abuse Act, and California Penal Code section 631 (wiretapping). The gravamen of the complaint is that ChatGPT allegedly accessed users’ platform access and intercepted their private information without their knowledge or consent. Motions to dismiss and to compel arbitration are pending.

Chabon v. OpenAI (N.D. Cal. 2023)

Complaint filed September 9, 2023. Authors allege that OpenAI infringed copyrights while training ChatGPT, and that ChatGPT is itself an unauthorized derivative work. They also assert claims of DMCA violations, unfair competition, negligence and unjust enrichment. The case has been consolidated with Tremblay v. OpenAI, and the cases are now captioned In re OpenAI ChatGPT Litigation.

Chabon v. Meta Platforms (N.D. Cal. 2023)

Complaint filed September 12, 2023. Authors assert copyright infringement claims against Meta, alleging that Meta trained its AI using their works and that the AI model itself is an unauthorized derivative work. The authors also assert claims for DMCA violations, unfair competition, negligence, and unjust enrichment. In November, 2023, the court issued an Order dismissing all claims except the claim of unauthorized copying in the course of training the AI. The court described the claim that an AI model trained on a work is a derivative of that work as “nonsensical.”

Authors Guild v. OpenAI, Microsoft, et al. (S.D.N.Y. 2023)

Complaint filed September 19, 1023. Book and fiction writers filed a complaint for copyright infringement in connection with defendants’ training AI on copies of their works without permission. A motion to dismiss has been filed.

Huckabee v. Bloomberg, Meta Platforms, Microsoft, and EleutherAI Institute (S.D.N.Y. 2023)

Complaint filed October 17, 2023. Political figure Mike Huckabee and others allege that the defendants trained AI tools on their works without permission when they used Books3, a text dataset compiled by developers; that their tools are themselves unauthorized derivative works; and that every output of their tools is an infringing derivative work.  Claims against EleutherAI have been voluntarily dismissed. Claims against Meta and Microsoft have been transferred to the Northern District of California. Bloomberg is expected to file a motion to dismiss soon.

Huckabee v. Meta Platforms and Microsoft (N.D. Cal. 2023)

Complaint filed October 17, 2023. Political figure Mike Huckabee and others allege that the defendants trained AI tools on their works without permission when they used Books3, a text dataset compiled by developers; that their tools are themselves unauthorized derivative works; and that every output of their tools is an infringing derivative work. Plaintiffs have filed an amended complaint. Plaintiffs have stipulated to dismissal of claims against Microsoft without prejudice.

Concord Music Group v. Anthropic (M.D. Tenn. 2023)

Complaint filed October 18, 2023. Music publishers claim that Anthropic infringed publisher-owned copyrights in song lyrics when they allegedly were copied as part of an AI training process (Claude) and when lyrics were reproduced and distributed in response to prompts. They have also made claims of contributory and vicarious infringement. Motions to dismiss and for a preliminary injunction are pending.

Alter v. OpenAI and Microsoft (S.D.N.Y. 2023)

Complaint filed November 21, 2023. Nonfiction author alleges claims of copyright infringement and contributory copyright infringement against OpenAI and Microsoft, alleging that reproducing copies of their works in datasets used to train AI infringed copyrights. The court has ordered consolidation of Author’s Guild (23-cv-8292) and Alter (23-cv-10211). On February 12,2024, plaintiffs in other cases filed a motion to intervene and dismiss.

New York Times v. Microsoft and OpenAI (S.D.N.Y. 2023)

Complaint filed December 27, 2023. The New York Times alleges that their news stories were used to train AI without a license or permission, in violation of their exclusive rights of reproduction and public display, as copyright owners. The complaint also alleges vicarious and contributory copyright infringement, DMCA violations, unfair competition, and trademark dilution. The Times seeks damages, an injunction against further infringing conduct, and a Section 503(b) order for the destruction of “all GPT or other LLM models and training sets that incorporate Times Works.” On February 23, 2024, plaintiffs in other cases filed a motion to intervene and dismiss this case.  

Basbanes and Ngagoyeanes v. Microsoft and OpenAI (S.D.N.Y. 2024)

Complaint filed January 5, 2024. Nonfiction authors assert copyright claims against Microsoft and OpenAI. On February 6, 2024, the court consolidated this case with Authors Guild (23-cv-08292) and Alter v. Open AI (23-cv-10211), for pretrial purposes.  

Caveat

This list is not exhaustive. There may be other cases involving AI that are not included here. For a discussion of bias issues in Google’s Gemini, have a look at Scraping Bias on Medium.com.

Nontransformative Nuge

A reversal in the 4th Circuit Court demonstrates the impact the Supreme Court’s decision in Andy Warhol Foundation for the Arts v. Goldsmith is already having on the application of copyright fair use doctrine in federal courts. Philpot v. Independent Journal Review,

A reversal in the 4th Circuit Court demonstrates the impact the Supreme Court’s decision in Andy Warhol Foundation for the Arts v. Goldsmith is already having on the application of copyright fair use doctrine in federal courts.

Philpot v. Independent Journal Review, No. 21-2021 (4th Circ., Feb. 6, 2024)

Philpot, a concert photographer, registered his photograph of Ted Nugent as part of a group of unpublished works. Prior to registration, he entered into a license agreement giving AXS TV the right to inspect his photographs for the purpose of selecting ones to curate. The agreement provided that the license would become effective upon delivery of photographs for inspection. After registration, Philpot delivered a set of photographs, including the Nugent photograph, to AXS TV. He also published the Nugent photograph to Wikimedia Commons under a Creative Commons (“CC”) license. The CC license allows free use on the condition that attribution is given. LJR published an article called “15 Signs Your Daddy Was a Conservative.” Sign #5 was He hearts the Nuge. LJR used Philpot’s photograph of Ted Nugent as an illustration for the article, without providing an attribution of credit to Philpot.

Philpot sued IJR for copyright infringement.  IJR asserted two defenses: (1) invalid copyright registration; and (2) fair use. The trial court did not decide whether the registration was valid or not, but it granted summary judgment for IJR based on its opinion that the news service’s publication of the photograph was fair use. The Fourth Circuit Court of Appeals reversed, ruling in Philpot’s favor on both issues. The Court held that the copyright registration was valid and that publication of the photograph without permission was not fair use.

The copyright registration

Published and unpublished works cannot be registered together. Including a published work in an application for registration of a group of unpublished works is an inaccuracy that might invalidate the registration, if the applicant was aware of the inaccuracy at the time of applying. Cf. Unicolors v. H&M Hennes & Mauritz, 595 U.S. 178 (2022). LJR argued that Philpot’s pre-registration agreement to send photographs to AJX TV to inspect for possible curation constituted “publication” of them so characterizing them as “unpublished” in the registration application was an inaccuracy known to Philpot.

17 U.S.C. § 101 defines publication as “the distribution of copies . . . to the public” or “offering to distribute copies . . . to a group of persons for purposes of further distribution . . . or public display.” The Court of Appeals held that merely entering into an agreement to furnish copies to a distributor for possible curation does not come within that definition. Sending copies to a limited class of people without concomitantly granting an unrestricted right to further distribute them to the public does not amount to “publication.”

Philpot’s arrangement with AXS TV is analogous to an author submitting a manuscript to a publisher for review for possible future distribution to the public. The U.S. Copyright Office has addressed this. “Sending copies of a manuscript to prospective publishers in an effort to secure a book contract does not [constitute publication].” U.S. Copyright Office, Compendium of U.S. Copyright Office Practices § 1905.1 (3d ed. 2021). Philpot had provided copies of his work for the limited purpose of examination, without a present grant of a right of further distribution. Therefore, the photographs were, in fact, unpublished at the time of the application for registration. Since no inaccuracy existed, the registration was valid.

Fair use

The Court applied the four-factor test for fair use set out in 17 U.S.C. § 107.

(1) Purpose and character of the use. Citing Andy Warhol Found. For the Visual Arts v. Goldsmith, 598 U.S. 508 , 527–33 (2023), the Court held that when, as here, a use is neither transformative nor noncommercial, this factor weighs against a fair use determination. LJR used the photograph for the same purpose as Philpot intended to use it (as a depiction of Mr. Nugent), and it was a commercial purpose.

(2) Nature of the work. Photographs taken by humans are acts of creative expression that receive what courts have described as “thick” copyright protection.” Therefore, this factor weighed against a fair use determination.

(3) Amount and substantiality of the portion used. Since all of the expressive features of the work were used, this factor also weighed against a fair use determination.

(4) Effect on the market for the work. Finally, the Court determined that allowing free use of a copyrighted work for commercial purposes without the copyright owner’s permission could potentially have a negative impact on the author’s market for the work. Therefore, this factor, too, weighed against a fair use determination.

Since all four factors weighed against a fair use determination, the Court reversed the trial court’s grant of summary judgment to IJR and remanded the case for further proceedings.

Conclusion

This decision demonstrates the impact the Warhol decision is having on copyright fair use analysis in the courts. Previously, courts had been interpreting transformativeness very broadly. In many cases, they were ending fair use inquiry as soon as some sort of transformative use could be articulated. As the Court of Appeals decision in this case illustrates, trial courts now need to alter their approach in two ways: (1) They need to return to considering all four fair use factors rather than ending the inquiry upon a defendant’s articulation of some “transformative use;” and (2) They need to apply a much narrower definition of transformativeness than they have been. If both the original work and an unauthorized reproduction of it are used for the purpose of depicting a particular person or scene (as distinguished from parodying or commenting on a work, for example), for commercial gain, then it would no longer appear to be prudent to count on the first of the four fair use factors supporting a fair use determination.

Photo: Photograph published in a July, 1848 edition of L’Illustration. Believed to be the first instance of photojournalism, it is now in the public domain.