Copyright And The Challenge of Large Language Models

by Lee Gesmer | Jul 1, 2024 | Copyright

“AI models are what’s known in computer science as black boxes: You can see what goes in and what comes out; what happens in between is a mystery.”

Trust but Verify: Peeking Inside the “Black Box” of Machine Learning

In December 2023, The New York Times filed a landmark lawsuit against OpenAI and Microsoft, alleging copyright infringement. This case, along with a number of similar cases filed against AI companies, brings to the forefront a fundamental challenge in applying traditional copyright law to a revolutionary technology: Large Language Models (LLMs). More than any copyright case that precedes them, these cases grapple with a form of alleged infringement that defies conventional legal analysis.

This article, the first in a three-part series, aims to explain how AI companies acquire vast amounts of training data and what happens to this data as it is processed into “artificial intelligence.” By understanding these technical aspects of LLMs, we can better appreciate the nuanced legal questions that arise when this technology intersects with copyright law, and why traditional methods of proving copyright infringement may be inadequate in this new technological landscape.

Disclaimer: I’m not a computer or AI scientist. However, neither are the judges and juries that will be asked to apply copyright law to this technology. It’s unlikely that they will go much beyond the level of detail I’ve used here.

What are Large Language Models (LLMs) and How Do They Work?

Large Language Models, or LLMs, are gargantuan AI systems that use a vast corpus of training data and billions of parameters. They are designed to understand, generate, and manipulate human language. They learn patterns from the data, allowing them to perform a wide range of language tasks with remarkable fluency. Their inner workings are fundamentally different from any previous technology that has been the subject of copyright litigation, including traditional computer software.

At their core, LLMs operate on a principle of prediction. Given a sequence of words an LLM attempts to predict the most likely next word. This process is repeated, with each prediction informing the next, allowing the model to generate coherent text, answer questions, or perform other language-related tasks.

The “large” in Large Language Models refers to the enormous number of parameters these models can contain – sometimes in the trillions. These parameters are essentially the model’s learned patterns and relationships, fine-tuned through exposure to massive amounts of text data. The larger the dataset, the better the AI model.

AI companies employ various methods to acquire this data – 

– Web scraping and crawling. One of the primary methods of data acquisition is web scraping – the automated process of extracting data from websites. AI companies deploy sophisticated crawlers that systematically browse the internet, copying text from millions of web pages. This method allows for the collection of diverse, up-to-date information but raises questions about the use of copyrighted material without explicit permission.

– Partnerships and licensing agreements. Some companies enter into partnerships or licensing agreements to access high-quality, curated datasets. For instance, OpenAI has partnered with organizations like the Associated Press to use its news archives for training purposes.

– Public datasets and academic corpuses. Many LLMs are trained, at least in part, on publicly available datasets and academic text collections. These might include Project Gutenberg’s collection of public domain books, scientific paper repositories, or curated datasets like the Common Crawl corpus.

– User-generated content. Platforms that interact directly with users, such as ChatGPT, can potentially use the conversations and inputs from users to further train and refine their models. This practice raises privacy concerns and questions about the ownership of user-contributed data.

In the context of the New York Times lawsuit, it’s worth noting that OpenAI, like many AI companies, has not publicly disclosed the full extent of its training data sources. However, it’s widely believed that the company uses a combination of publicly available web content, licensed datasets, and partnerships to build its training corpus. The lawsuit alleges that this corpus includes copyrighted New York Times articles, obtained without permission or compensation.

Once acquired, the raw data undergoes several processing steps before it can be used to train an LLM – 

– Data preprocessing and cleaning. The first step involves cleaning the raw data. This includes removing irrelevant information, correcting errors, and standardizing the format. This may involve stripping away HTML tags, removing advertisements, or filtering out low-quality content.

– Tokenization and encoding. Next, the text is broken down into smaller units called tokens. These might be words, parts of words, or even individual characters. Each token is then converted into a numerical representation that the AI can process. This step is crucial as it determines how the model will interpret and generate language.

The Training Process: How Machines “Learn” From Data

During training, the LLM is exposed to this preprocessed data, learning to predict patterns and relationships between tokens. This is an iterative process where the model makes predictions, compares them to the actual data, and adjusts its internal parameters to improve accuracy. This process, known as “backpropagation,” is repeated billions of times across the entire dataset. In a large LLM this can take months, operating 24/7 on a massive system of graphics processing chips.

The Invisible Transformation: From Text to Numbers

For purposes of copyright law, here’s the crux of the matter – after this process, the original text no longer exists in any recognizable form within the LLM. The model becomes a vast sea of numbers, with no direct correspondence to the original text. This transformation creates a fundamental challenge for copyright law – 

– No Side-by-Side Comparison: In traditional copyright cases, courts rely heavily on comparing the original work side-by-side with the allegedly infringing material. With LLMs, this is impossible. You can’t “read” an LLM or print it out for comparison.

– Black Box Nature: The internal workings of LLMs are often referred to as a “black box.” Even the developers may not fully understand how the model arrives at its outputs.

– Dynamic Generation: LLMs don’t store and retrieve text; they generate it dynamically based on learned patterns. This means that any similarity to copyrighted material in the output is a result of statistical prediction, not direct copying.

– Distributed Information: Information from any single source is distributed across countless parameters in the model, making it impossible to isolate the influence of any particular work.

Legal Implications

This technological reality creates several unprecedented challenges for copyright law – 

– Proving Infringement: How can a plaintiff prove infringement when the allegedly infringing material can’t be directly observed or compared?

– Fair Use Analysis: Traditional fair use factors, such as the amount and substantiality of the portion used, become difficult to apply when the “portion used” is transformed beyond recognition.

– Substantial Similarity: The legal test of “substantial similarity” between works becomes almost meaningless in the context of LLMs.

– Expert Testimony: Courts will likely have to rely heavily on expert testimony to understand the technology, but even experts may struggle to definitively prove or disprove infringement.

For all of these reasons, to prove copyright infringement plaintiffs such as the New York Times may have to rely on the “intermediate” copies that are used to start the training process, rather than the “end-product” LLMs that are accessed by users. And this, as we’ll see in my next post in this series, may favor OpenAI and other AI companies. 

Conclusion

The NYT v. OpenAI case highlights a fundamental mismatch between traditional copyright law and the reality of LLM technology. As I move forward in this series, I’ll explore the legal theories and potential new frameworks that might be necessary to address these challenges. The outcome of this case could reshape our understanding of copyright in the digital age, potentially requiring new legal tests and standards that can account for the invisible, transformed nature of information within AI systems.

Part 2 in this series will focus on the legal issues around the “input problem” of using copyrighted material for training. Part 3 will look at the “output problem” of AI-generated content that may copy or resemble copyrighted works, including what the AI industry calls “memorization.” As we’ll see, each of these issues presents its own unique challenges in the context of a technology that defies traditional legal analysis.

Secondary Liability and Sony v. Cox

by Lee Gesmer | May 26, 2024 | Copyright

Copyright secondary liability can be difficult to wrap your head around. This judge-made copyright doctrine allows copyright owners to seek damages from organizations that do not themselves engage in copyright infringement, but rather facilitate the infringing behavior of others. Often the target of these cases are internet service providers, or “ISPs.”

Secondary liability has three separate prongs, “contributory” and “vicarious” infringement, and “inducement.” The third prong – inducement – is important but seen infrequently. For the elements of this doctrine see my article here. 

Here’s how I outlined the elements of contributory and vicarious liability when I was teaching CopyrightX:

These copyright rules were the key issue in the Fourth Circuit’s recent blockbuster decision in Sony v. Cox Communications (4th Cir. Feb. 20, 2024).

In a highly anticipated ruling the court reversed a $1 billion jury verdict against Cox for vicarious liability but affirmed the finding of contributory infringement. The decision is a significant development in the evolving landscape of ISP liability for copyright infringement.

Case Background

Cox Communications is a large telecommunications conglomerate based in Atlanta. In addition to providing cable television and phone services it acts as an internet service provider – an “ISP” – to millions of subscribers. 

The case began when Sony and a coalition of record labels and music publishers sued Cox, arguing that the ISP should be held secondarily liable for the infringing activities of its subscribers. The plaintiffs alleged that Cox users employed peer-to-peer file-sharing platforms to illegally download and share a vast trove of copyrighted music, and that Cox fell short in its efforts to control this rampant infringement.

A jury found Cox liable under both contributory and vicarious infringement theories, levying a jaw-dropping $1 billion in statutory damages – $99,830.29 for each of the 10,017 infringed works. Cox challenged the verdict on multiple fronts, contesting the sufficiency of the evidence and the reasonableness of the damages award.

The Fourth Circuit Opinion

On appeal, the Fourth Circuit dissected the two theories of secondary liability, arriving at divergent conclusions. The court sided with Cox on the issue of vicarious liability, finding that the plaintiffs failed to establish that Cox reaped a direct financial benefit from its subscribers’ infringing conduct. Central to this determination was Cox’s flat-fee pricing model, which remained constant irrespective of whether subscribers engaged in infringing or non-infringing activities. The mere fact that Cox opted not to terminate certain repeat infringers, ostensibly to maintain subscription revenue, was deemed insufficient to prove Cox directly profited from the infringement itself.

However, the court took a different stance on contributory infringement, upholding the jury’s finding that Cox materially contributed to known infringement on its network. The court was unconvinced by Cox’s assertions that general awareness of infringement was inadequate, or that a level of intent tantamount to aiding and abetting was necessary for liability to attach. Instead, the court articulated that supplying a service with the knowledge that the recipient is highly likely to exploit it for infringing purposes meets the threshold for contributory liability.

Given the lack of differentiation between the two liability theories in the jury’s damages award, coupled with the potential influence of the now-overturned vicarious liability finding on the damages calculation, the court vacated the entire award. The case now returns to the lower court for a new trial, solely to determine the appropriate measure of statutory damages for contributory infringement.

Relationship to the DMCA

This article’s header graphic illustrates the relationship between the secondary liability doctrines and the protection of the Digital Millennium Copyright Act (DMCA), Section 512(c) of the Copyright Act. As the graphic reflects, all three theories of secondary liability lie outside the DMCA’s safe harbor protection for third-party copyright infringement. The DMCA requires that a defendant satisfy multiple safe harbor conditions (See my 2017 article – Mavrix v. LiveJournal: The Incredible Shrinking DMC for more on this). If a plaintiff can establish the elements of any one of the three theories of secondary liability the defendant will violate one or more safe harbor conditions and lose DMCA protection.

Implications

The court’s decision signals a notable shift in the contours of vicarious liability for ISPs in the context of copyright infringement. By requiring a causal nexus between the defendant’s financial gain and the infringing acts themselves, the court has raised the bar for plaintiffs seeking to prevail on this theory.

The ruling underscores that simply profiting from a service that may be used for both infringing and non-infringing ends is insufficient; instead, plaintiffs must demonstrate a more direct and meaningful link between the ISP’s revenue and the specific acts of infringement. This might entail evidence of premium fees for access to infringing content or a discernible correlation between the volume of infringement and subscriber growth or retention.

While Cox may take solace in the reversal of the $1 billion vicarious liability verdict, the specter of substantial contributory infringement damages looms large as the case heads back for a retrial.

For ISPs, the ruling serves as a warning to reevaluate and fortify their repeat infringer policies, ensuring they go beyond cosmetic compliance with the DMCA’s safe harbor provisions. Proactive monitoring, prompt responsiveness to specific infringement notices, and decisive action against recalcitrant offenders will be key to mitigating liability risks.

On the other side of the equation, copyright holders may need to recalibrate their enforcement strategies, recognizing the heightened evidentiary burden for establishing vicarious liability. While the contributory infringement pathway remains viable, particularly against ISPs that display willful blindness or tacit encouragement of infringement, the Sony v. Cox decision underscores the importance of marshaling compelling evidence of direct financial benefit to support vicarious liability claims.

As this case enters its next phase, the copyright and technology communities will be focused on the outcome of the damages retrial. Regardless of the ultimate figure, the Fourth Circuit’s decision has already left a mark on the evolving landscape of online copyright enforcement.

Header image is published under the Creative Commons Attribution 4.0 License.

Supreme Court Allows Copyright Damages Beyond 3 Years – But Leaves Key Question Open

by Lee Gesmer | May 17, 2024 | Copyright

Many aspects of copyright law are obscure and surprising, even to lawyers familiar with copyright’s peculiarities. An example of this is copyright law’s three-year statute of limitations. 

The Copyright Act states that “no civil action shall be maintained under the provisions of this title unless it is commenced within three years after the claim accrued.” 17 U. S. C. §507(b). In the world of copyright practitioners this is understood to mean that so long as a copyright remains in effect and infringements continue, an owner’s rights are not barred by the statute of limitations. However, they may be limited to damages that accrued in the three years before the owner files suit. This is described variously as a “three-year look-back,” a “rolling limitations period” or the “separate-accrual rule.”

This is what allowed Randy Wolfe’s estate to sue Led Zeppelin in 2014 for an alleged infringement that began in 1971.

However, there is a nuance to this doctrine – what if the copyright owner isn’t aware of the infringement? Is the owner still limited to damages accrued in the three years before he files suit? 

That is the scenario the Supreme Court addressed in Warner Chappell Music, Inc. v. Nealy (May 9, 2024). 

Background Facts

Songwriter Sherman Nealy sued Warner Chappell in 2018 for infringing his music copyrights going back to 2008. Warner responded that under the “three year look-back” rule Nealy’s damages were limited to three years before he filed suit. Nealy argued that his damages period should extend back to 2008, since his claims were timely under the “discovery rule” – he was in prison during much of this period and only learned of the infringements in 2016.

Nealy lost on this issue in the district court, which limited his damages to the infringer’s profits during the 3 years before he filed suit. The 11th Circuit reversed, holding that Nealy could recover damages beyond 3 years if his claims were timely – meaning that the case was filed within three years of when Nealy discovered the infringement. 

The Supreme Court Decision

The Supreme Court affirmed the 11th Circuit and resolved a circuit split, holding: 

1 – The Copyright Act’s 3-year statute of limitations governs when a claim must be filed, not how far back damages can go.

2 – If a claim is timely, the plaintiff can recover damages for all infringements, even those occurring more than 3 years before suit. The Copyright Act places no separate time limit on damages.

However, lurking within this ruling is another copyright law doctrine that the Court did not address that could render its ruling in Nealy moot – that is the proper application of the “discovery rule” under the Copyright Act. Under the discovery rule a claim accrues when “the plaintiff discovers, or with due diligence should have discovered” the infringement. (Nealy, Slip Op. p. 2). Competing with this is the less liberal “occurrence” rule, which holds that, in the absence of fraud or concealment, the clock starts running when the infringement occurs. Under the discovery rule Nealy would be able to recover damages back to 2008. Under the occurrence rule his damages would be limited to the three years before he filed suit, since he does not allege fraud or concealment. 

However, the question of which rule applies under the Copyright Act has never been addressed by the Supreme Court, and is itself the subject of a circuit split. The Court assumed, without deciding and solely for purposes of deciding the issue before it, that the discovery rule does apply to copyright claims. If the discovery rule applies Nealy has a claim to retroactive damages beyond three years. If it does not, Nealy’s damages would be limited to the three years before he filed suit.

Justice Gorsuch, joined by Justices Thomas and Alito, focused on this in his dissent, arguing the Court should not have decided the issue when the “discovery vs. occurrence” issue has not been addressed:

The Court discusses how a discovery rule of accrual should operate under the Copyright Act. But in doing so it sidesteps the logically antecedent question whether the Act has room for such a rule. Rather than address that question, the Court takes care to emphasize that its resolution must await a future case. The trouble is, the Act almost certainly does not tolerate a discovery rule. And that fact promises soon enough to make anything we might say today about the rule’s operational details a dead letter.

Clearly, in the view of at least three justices, if and when the discovery vs. occurrence rule issue comes before the Court it could decide against the discovery rule in copyright cases, rendering its decision on damages in the Nealy case, and cases like it, moot.

State of the Law Today

What does this all boil down to? Here are the rules as they exist today –

– A copyright owner has been aware of an infringing musical work for 20 years. She finally sues the infringer. Her damages are limited by the three year damages bar. They may be limited even further based on the laches doctrine.

– A copyright owner has been meditating alone in a cave in Tibet for 20 years. She’s had no access to information from the outside world. Upon her return she discovers that someone has been infringing her literary work for the last 20 years. Depending on whether the federal circuit applies the discovery or the occurrence rule, she may recover damages for the entire 20 period, or just the preceding three years. Her lawyers should do some careful forum shopping.

– A copyright owner discovers someone has secretly been infringing her copyright in computer source code for 20 years. The source code was non-public, and therefore the infringement was concealed. She may recover damages for the full 20 year period.

Implications

The decision is a win for copyright plaintiffs, allowing them to reach back and get damages beyond 3 years – assuming their claims are timely and they are in a circuit that apples the discovery rule. But the Court left the door open to decide the more important question of whether the discovery rule applies to the Copyright Act’s statute of limitations at all. If not, the window for both filing claims and recovering damages will shrink. When this issue will reach the Supreme Court is uncertain. However, the Court has the opportunity to take it up as soon as next term. See Hearst Newspapers, LLC v. Martinelli, No. 23-474 (U.S. petition for cert. filed Nov. 2, 2023). In the meantime, the outer boundary of damages is limited only by the discovery rule (if it apples), not any separate damages bar. Plaintiffs with older claims should take note, as should potential defendants doing due diligence on liability exposure.

Update: On May 20, 2024, the Supreme Court of the United States denied the petition for certiorari in Hearst Newspapers, L.L.C. v. Martinelli, thereby declining to decide whether the discovery rule applies to copyright infringement claims and leaving the rule intact. 

Header image attribution: Resource by Nick Youngson CC BY-SA 3.0 Pix4free

Court Denies Copyright Protection to AI Generated Artwork

by Lee Gesmer | Aug 28, 2023 | Copyright

In an previous post I focused on the AI “output” issue – who owns an AI model’s output? (Artificial Intelligence May Result In Human Extinction, But In the Meantime There’s a Lot of Lawyering To Be Done).  I noted that this issue was pending in a lawsuit before the Federal District Court for the District of Columbia (Thaler v. Perlmutter).

The decision in this case was issued by Judge Beryl A. Howell on August 18, 2023. In her ruling Judge Howell made it clear that a creation born out of an artificial intelligence system cannot be copyrighted due to the lack of human creativity, the “sine qua non at the core of copyrightability.”

Background

In 2019 Stephen Thaler filed an unusual copyright application. Instead of a traditional artwork, the piece – titled “A Recent Entrance to Paradise” (the image appears at the top of this post) – identified an unusual ‘creator’ – the “Creativity Machine.” The Creativity Machine is an AI system invented by Thaler. In his application for registration Thaler informed the Copyright Office that the work was “created autonomously by machine,“ and his claim to the copyright was based on the fact of his “ownership of the machine.“

The Copyright Office, however, didn’t see it his way. Its position is that that copyright protections are reserved exclusively for works born from human ingenuity. See Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence. On this basis it declined Thaler’s application.

Judge Howell’s Decision

Oscar Wilde photo in Burrow-Giles case

On appeal to the district court Judge Howell acknowledged that copyright law is “malleable enough to cover works created with or involving technologies developed long after traditional media.” A prime example of this is the U.S. Supreme Court’s 1884 decision in Burrow-Giles Lithographic Co. v. Sarony, holding that a photograph of Oscar Wilde was copyrightable despite use of a camera, since the camera was used to give “visible expression” to “ideas in the mind of the author.” However, the rationale in this case didn’t go far enough for Judge Howell. Her decision emphasized the foundational principle of copyright: human creativity – 

The act of human creation—and how to best encourage human individuals to engage in that creation, and thereby promote science and the useful arts—[has been] central to American copyright from its very inception. Non-human actors need no incentivization with the promise of exclusive rights under United States law, and copyright was therefore not designed to reach them.

The Copyright Act itself leans heavily toward human-centric creation, with previous court decisions strengthening the court’s perspective. The case that has received much of the attention on this topic is Naruto v. Slater, where a photograph, though artistically noteworthy, wasn’t given copyright protection because it was taken by a monkey, not a human – “all animals, since they are not human,” lack standing under the Copyright Act. 

Attempting to navigate this legal maze, on appeal Thaler presented a fresh angle. He argued that as the mastermind behind the AI, providing it with direction and instructions, he should be considered the rightful human author. But this theory had not been asserted in his original application for registration, and therefore was dismissed by the court: “here, plaintiff informed the register that the work was ‘created autonomously by machine,’ and his claim to the copyright was only based on the fact of his ‘ownership of the machine.'” Therefore, the court limited Thaler’s appeal to the question of whether a work generated autonomously by a computer system is eligible for copyright, and held that it was not.

Navigating Uncharted Waters: Future Implications

9th Cir in Naruto case: Monkey selfie not copyright-protected

The Thaler case sets a precedent: a creation made entirely by an AI, without human intervention, remains outside the protective bounds of the Copyright Act – at least for now. Not surprisingly, Thaler has announced that he will appeal this ruling to the D.C. Circuit. Onward and upward.

Moreover, this case leaves unaddressed a myriad of yet-to-be-answered questions:

• At what point, and to what extent – if at all – does human interaction with AI validate a creation as human-made?
• How do we gauge the originality of AI creations when these AI systems might have been trained using pre-existing works?
• Should the current structure of copyright be reformed to support and foster AI-involved creations?

These questions remain tantalizingly open, awaiting future exploration and legal interpretation. The ongoing debate about AI’s role in the world of creativity and copyright is just beginning.

Thaler v. Perlmutter (D. C. August 18, 2023)

Taylor’s Version and the Two Copyrights

by Lee Gesmer | Jul 15, 2023 | Copyright

A couple of people have asked me about the legal story behind Taylor Swift’s re-recording of her earlier albums. 

Great question. In fact, she has re-recorded three of them. 

This unusual story is a perfect “music copyright” teaching moment.

Why The Re-Recordings?

The background is a bit convoluted, but it arises out of an ugly split between Swift and her first recording company, Big Machine Records. Following the split Swift began releasing her re-recorded songs, Fearless (Taylor’s Version) and Red (Taylor’s Version) in 2021 and Speak Now (Taylor’s Version) in 2023.

Why did she re-record the songs on these albums? The gory details are discussed under the link above, but after the falling out with  Big Machine, Swift decided to re-record the songs owned by it, apparently with the intention of diverting sales from her former recording company. 

Swift’s popularity and financial resources allow her to do something few other artists could hope to undertake.

Copyright Law and Music

There is an important aspect of copyright law at the heart of what happened here. Every musical recording potentially has two copyrights – one in the musical work and one in each recording of the work. The musical work is the composition – the chords, melody and lyrics. Swift penned the songs on these three albums and as the author, retained ownership of these musical works. However, she assigned the recordings or “masters” to her recording company. Although she might earn royalties based on the sales and performances of these masters, she doesn’t own the copyright for them.  

By not also assigning ownership of the musical works represented by the songs on the three albums, Swift retained what the music industry refers to as “publishing rights,” as in “hey, I own the publishing for this song, right?” Swift is therefore free to re-record them, as she has now done in the three “Taylor’s Version” albums. 

Further Intricacies and Questions

It’s likely that there’s more to this story than has been revealed to the public. For instance, a contract between Swift and Big Machine may have temporarily delayed Swift from re-recording her songs. However, that’s more about contract law than copyright. The music industry, often a confusing maze, juggles both copyrights and contracts.

The extent to which Swift is getting her hoped-for revenge is unknown – we don’t know the extent to which the re-recordings are cutting into sales of the original masters. And, no doubt there are many other legal complications that have not been made public. For example, assume a movie producer wants a “synchronization license” (a “sync” license) to use one of these recordings with a movie or TV show. The producer needs a license to both the master and the musical work. I can imagine Taylor Swift saying, “if you want a license to the musical work you need to license the new master from me as well.” This would cut out the owner of the first recording, and no doubt lead to threats of contractual interference. But is it legal? It probably is. 

When I introduced the distinction between the copyrights in musical works and masters above, I said that “every musical recording potentially has two copyrights.” Why did I say “potentially”?

An example will illustrate why. Assume that in 2023 a symphony orchestra records and releases a performance of Antonín Dvořák’s NewWorld Symphony, composed in 1893. The copyright in the musical work has expired. Anyone is free to record this work. However, a new copyright applies to the new recording and will last for decades. Thus, only one copyright – the copyright in the master – exists in this scenario.

If you’re interested in the drama between Taylor Swift and her former record company, this Wikipedia  entry has most of  it.

Image credit: Eva Rinaldi https://commons.wikimedia.org/wiki/File:Taylor_Swift_%286966830273%29.jpg

The Impact of SAS Institute, Inc. v. World Programming Ltd. on Non-Literal Software Copyright Infringement

by Lee Gesmer | Apr 21, 2023 | Copyright

Software copyright is an important area of copyright law. However, it has proven devilishly difficult for the courts to apply. As the Second Circuit observed 30 years ago, trying to apply copyright law to software is often an “attempt to fit the proverbial square peg in a round hole.” Judges know this – I’ll never forget the time that Massachusetts Federal District Court Judge Rya Zobel, during an initial case conference in a copyright case, looked me in the eye and said, “we aren’t going to have to compare source codes in this case, are we Mr. Gesmer?” (We didn’t, the case settled soon afterwards).

The Court of Appeals for the Federal Circuit (the CAFC) has grappled with this challenge, most notably in its two controversial decisions in Oracle v. Google. (2014, 2018).

Now the CAFC has issued an important decision in SAS Institute, Inc. v. World Programming Limited (April 6, 2023; Newman dissenting). The issue in this case is one that I encountered in a copyright suit in Boston, so it’s of particular interest to me. More on that below. 

SAS Institute and World Programming

SAS Institute is a successful software company. Its annual revenues exceed $3 billion, and it has more than 12,000 employees. Its statistical analysis software — the “SAS System” – is used in 3,000+ companies worldwide. 

Success attracts imitation, and World Programming (now part of Altair) developed a “clone” of the SAS Software. SAS didn’t react kindly to the competition – it has conducted a more-than 10 year, multi-nation legal challenge, suing World Programming once in England and twice in the United States. 

What makes SAS’s most recent copyright case against World Programming unusual is the subject matter. Most software copyright litigation involves the “literal elements” of computer programs – the “source” and “object” code – essentially the “written words” or the machine code (ones and zeros) of the software.

“Non-literal” Copyright Infringement

SAS v. World Programming, however, involved the “non-literal” elements of SAS’s system. The courts define “non-literal elements” as the structure, sequence, and organization and the user interface of software. Basically, anything other than the computer code. SAS alleged that World Programming illegally copied input syntax formats and output design styles – non-literal components of the SAS System.

The idea that non-literal components of a software program can be protected by copyright has been acknowledged since the 1980s. For the last 30 years most courts have followed the “abstraction-filtration-comparison” test (AFC test) established in the 1992 Second Circuit decision in Altai v. Computer Associates. The AFC test requires the court to (1) break a software program into its constituent parts (abstraction), (2) filter out unprotectable elements (filtration) and (3) compare the remaining protectable elements to the allegedly infringing work (comparison).

If this sounds challenging to you, you are right. However, relatively few cases have actually had to undertake the real-world application of this test to the non-literal elements of a software program. And, where they have the plaintiff has almost always lost.

The District Court Case

SAS filed this case in the Eastern District of Texas. The district court judge proceeded to apply the Altai AFC test by conducting a hearing to “filter out” unprotectable elements of the SAS software. Examples of unprotected elements include ideas, facts, information in the public domain, merger material, scènes à faire and conventional display elements. Case law has established that abstraction and filtration (steps 1 and 2 of the AFC test) is performed by the judge, not the jury. 

The district court held what it termed a “copyrightability hearing” and implemented an alternating, burden-shifting framework in which SAS was required to prove a valid copyright and “factual copying.” The burden then shifted to defendant (World Programming) to prove that some or all of the copied material is unprotectable. The burden then shifted back to SAS to respond and persuade the court otherwise. 

Think of this as a tennis volley in which the ball crosses the net three times. 

SAS satisfied the first part of this test – it showed that it had a registered copyright, and that World Programming had copied some elements of the SAS System. However, World Programming responded with evidence that many of the non-literal components of the SAS System contained factual elements, elements that were not original to SAS or that were in the public domain, unprotected mathematical and method components, conventional display elements and merger elements. World Programming asserted that all of these components should be filtered out and excluded from step 3 of the AFC test – comparison of the two software programs. 

At that point, under the judge’s burden shifting approach, the burden fell on SAS to respond and address these defenses. 

Inexplicably, SAS failed to do this. The court stated – 

SAS has not attempted to show what World Programming pointed to as unprotectable is indeed entitled to protection. . . . Instead, when the burden shifted back to SAS, it was clear SAS had done no filtration; they simply repeated and repeated that the SAS System was “creative.” . . . SAS’s failures have raised the untenable specter of the Court taking copyright claims to trial without any filtered showing of protectable material within the asserted work. This is not a result that this Court can condone. These failures rest solely on SAS and the consequences of those failures necessarily rest upon SAS as well.

The district court then dismissed the case. SAS appealed to the Federal Circuit – a court that is notoriously pro-copyright. (See the two Oracle decisions linked to above). SAS likely planned for any appeal to go to the Federal Circuit by asserting patent infringement against World Programming and later dropping its patent claims. Nevertheless, that was enough to give the Federal Circuit jurisdiction over any appeal.

Appeal to the Federal Circuit

On appeal the central question was procedural: Was it SAS’s burden to prove that the copied elements were protectable, or was it World Programming’s burden to prove that they were not? In other words, the issue was who bears the burden of proving, as part of the filtration analysis, that the elements the defendant copied are unprotectable – the plaintiff (copyright owner) or the defendant (alleged infringer)?

The Federal Circuit was not impressed with SAS’s arguments on appeal. It noted that rather than participate in the steps required by the Altai AFC test, SAS “failed or refused” to identify the constituent elements of the SAS software that it claimed were protectable. Instead, it argued that its software was “creative” and that it had provided evidence that World Programming had engaged in “factual copying.” But it provided no evidence in relation to the “filtration” step under the 3-part Altai AFC test. 

The Federal Circuit found the trial court judge’s procedure to be appropriate: “a court may reasonably adopt an analysis to determine what the ‘core of protectable expression’ is to provide the jury with accurate elements to compare in its role of determining whether infringement has occurred.” The court concluded that SAS failed to “articulate a legally viable theory” and affirmed dismissal.

In other words, to continue the tennis analogy, SAS served the ball (showed that it had copyright registrations and that World Programming had copied some elements). World Programming returned the ball, introducing evidence that many of the elements SAS had identified were unprotected by copyright, and needed to be “filtered out” before the SAS and World Programming software programs were compared. However, SAS was unable to return that volley – “The district court found that SAS refused to engage in the filtration step and chose instead to simply argue that the SAS System was ‘creative.’”

20-20 Design v. Real View – Same Issue, No Controversy

While this is an important software copyright case and will be used defensively by copyright defendants in the future, it caught my attention for a second reason, which is that I dealt with the same issue in 20-20 Design v. Real View LLC, a copyright infringement case I tried to a jury in Boston in 2010. That case dealt with the graphical user interface of a software program – “nonliteral” elements of the software. Like World Programming in the SAS case, Real View allegedly created a “clone” program, but the cloning didn’t involve the source or object code, only parts of the graphical user interface.

Massachusetts Federal District Court Judge Patti Saris ordered 20-20 Design, the plaintiff/copyright owner, to identify the elements of its software that it claimed had been infringed. Unlike SAS, 20-20 Design complied. It provided a list of 60 elements, and the court held what Judge Saris called (by analogy) a “Markman”-style evidentiary hearing, which included evidence and testimony from experts on both sides. In effect, this was the “copyrightability hearing” held by the court in the SAS case. 

Judge Saris then issued a copyrightability decision holding that almost all of the items were not individually protectible. They could, however, be protected as a “compilation.” However, she ruled that as a “compilation,” the plaintiff-copyright owner was required to prove that the defendant’s software interface was “virtually identical” – a much more difficult standard to meet than the “substantial similarity” standard applied in most copyright litigation. 

(Humble brag: 20-20 Design was seeking damages of $2.8 million. However, the “virtually identical” standard proved to be its downfall. Without going into detail, suffice it to say that after a 10-day jury trial and post-trial motions the judge entered judgment for 20-20 Design against Real View (my client) in the amount of $4,200. (link)

When I read the decision in SAS v. World Programming I immediately related it to the 20-20 Design/Real View case, but I couldn’t recall how Judge Saris had allocated the burden-of-proof. When I refreshed my memory I realized why – the judge and the parties never discussed this issue. It seems that everyone assumed that the plaintiff-copyright holder (20-20 Design) had the burden of proof. After 20-20 identified the copied elements Real View argued that most of them should be filtered out and 20-20 Design (unlike SAS) responded with counter arguments. In other words, the ball went over the net three times, and the judge was able to apply the Altai AFC test and “filter” 20-20’s software before trial.

Thinking back on how smoothly this procedure went in my case, it’s difficult for me to imagine how SAS chose the strategy that cost it the World Programming case, unless this case was just an attempt to outspend a smaller competitor and drive it out of the market with litigation expenses. SAS is a multi-billion-dollar company. Its lawyers are highly experienced. Why SAS chose a case strategy that seemed doomed to failure is a bit of a mystery. One possibility is that SAS knew that if it identified the elements it would be forced into a copyright compilation theory that requires proof that the infringing work is “virtually identical” to plaintiff’s work, a burden that SAS believed it could not satisfy. Another is that it gambled that the Federal Circuit – which is notoriously protective of copyright owners – would see the law its way and reverse the district court. We will never know. 

Although it remains a mystery why SAS chose a case strategy that seemed destined to fail, the SAS v. World Programming decision has important implications for software copyright law. It clarifies the burden-shifting process and emphasizes the importance that the plaintiff be fully prepared to engage in the Altai AFC test’s filtration step.

Will SAS appeal this decision to the Supreme Court? Given the resources that SAS has dedicated to its litigation with World Programming over the last decade it seems likely that it will. While I view it as doubtful that the Supreme Court will hear this case, you never know.

SAS Institute v. World Programming (Fed. Cir. April 6, 2023)

Generative AI Images Struggle for Copyright Protection

by Lee Gesmer | Mar 6, 2023 | Copyright

There are a number of computer programs and websites that will allow you to create an image using artificial intelligence. One of them is Midjourney. You can see some of the Midjourney AI-generated art here. 

Kris Kashtanova used Midjourney’s generative AI tool to create a comic book titled Zarya of the Dawn. She submitted the work to the Copyright Office, seeking registration, and the Office issued the registration in September 2022. However according to the Copyright Office Ms. Kashtanova did not disclose that she used artificial intelligence to create Zarya.

Soon afterwards the Office became aware – via a reporter’s inquiry and social media posts – that Ms. Kashtanova had created the comic book using artificial intelligence. The Office reconsidered the registration and, after much correspondence and argumentation with Ms. Kashtanova’s attorneys, canceled the registration, concluding that:

. . . the images in the Work that were generated by the Midjourney technology are not the product of human authorship. Because the current registration for the Work does not disclaim its Midjourney-generated content, we intend to cancel the original certificate issued to Ms. Kashtanova and issue a new one covering only the expressive material that she created. 

Image from Zarya

This conclusion is the denouement in a lengthy letter from the Copyright Office analyzing the copyrightability of the images contained in the Zarya comic in detail in light of how Midjourney creates images. In correspondence with the Copyright Office Ms. Kashtanova argued that she had provided “hundreds or thousands of descriptive prompts” to Midjourney to generate “as perfect a rendition of her vision as possible.” However, based on how Midjourney creates images – essentially via a random mechanical process, notwithstanding the prompts of the human “mastermind” – the Copyright Office concluded that she was not the “author” of the resulting images for copyright purposes. The Copyright Office reasoned that “unlike other tools used by artists” (such as Adobe Photoshop), Midjourney generates images using prompts in an “unpredictable way.” “Because of the significant distance between what a user may direct Midjourney to create and the visual material Midjourney actually produces,” Ms. Kashtanova did not have enough control over the final images generated to be the “inventive or mastermind” behind the images.

Here are some takeaways from this decision.

First, artists using generative AI to create images should not assume that they own a copyright in the images.  At present the Copyright Office appears firmly committed to its position that they do not, and until there are court decisions to the contrary, or Congress amends the Copyright Act to accommodate these works, the better practice is to assume no protection.

Second, it may be possible to protect an AI-created work based on human modifications to the work. This was illustrated by the Zarya decision, where Ms. Kashtanova also sought registration for images that she created using Midjourney but altered post-production using Photoshop. With respect to one of these images the Copyright Office left open the possibility that copyrightable expression had been added, and therefore the image might receive registration. However, in these cases the burden will be on the human artist to establish that the human modifications or contributions reflect sufficient expression to receive protection. And, the scope of protection would likely be limited to the modifications, not the full image.

Image from Zarya

Third, this is a fast-moving area of law. Ms. Kashtanova – or any person or company denied registration – has the right to appeal the Copyright Office’s decision to a federal district court, from which the case may go on appeal to a circuit court, or even the Supreme Court. Whether Ms Kashtanova will take that action – or whether we will have to wait for another case – remains to be seen. A court – or Congress by amendment to the Copyright Act – could change the law on copyright protection of AI images.

Lastly, the Copyright Office’s reasoning on AI images is likely to extend to text as well. Thus, if a person uses a program such as ChatGPT to create a written work, it seems unlikely that the Copyright Office would accept it for purposes of registration. Despite the best efforts of the “prompt engineer,” the resulting output is likely to be too random to fall within the Copyright Office’s views of authorship.

Update: On March 10, 2023, less than a week after I published this post, the Copyright Office issued a “statement of policy to clarify its practices for examining and registering works that contain material generated by the use of artificial intelligence technology.” (link). Here is the heart of that policy statement: “In the case of works containing AI-generated material, the Office will consider whether the AI contributions are the result of ‘mechanical reproduction’ or instead of an author’s ‘own original mental conception, to which [the author] gave visible form.’ The answer will depend on the circumstances, particularly how the AI tool operates and how it was used to create the final work. This is necessarily a case-by-case inquiry.”

How this principle will be applied in practice remains to be seen.

Music, Copyright and the Performing Rights Organizations

by Lee Gesmer | Sep 3, 2022 | Copyright

With some exceptions, every public venue that plays popular music for its customers – concert venue, bar, restaurant, shopping mall or health club – needs to enter into a blanket license agreement with ASCAP, BMI and SESAC, the performing rights organizations (PROs) that pay public performance royalties to songwriters and publishers. 

Occasionally a club will fail to join a PRO, ignore warnings and be sued for copyright infringement. 

Here’s a current example in which a club did sign a license with ASCAP,  but allegedly failed to pay ASCAP the license fees.

Universal Music v. Calvin Theater

Two music publishers have sued Calvin Theater and its owner/manager, Eric Suher, for violating public performance rights in six compositions. Calvin Theater is a music venue in Northampton, Massachusetts, and the suit was filed in the Federal District Court for the District of Massachusetts. ASCAP manages the public performance rights for these songs.

We don’t know why Calvin Theater failed to pay ASCAP. However, the publishers’ claim is that because the venue is in breach the agreement is not in effect and the Theater has no copyright license.  

Based on the allegations in the complaint this case is a good opportunity to do a short tutorial on the intersection of the music industry and copyright law.

The Infringement Involves “Musical Compositions,” Not “Sound Recordings”

Every musical work can have two copyrights – the musical composition (melody, lyrics) and the sound recording. The owners can be different, and usually are – it’s common for a record company to own the copyright in the “master” sound recording and a publishing company to own rights in the musical composition. 

Calvin Theater involves the public performance of musical compositions – there is no allegation that sound recordings were illegally copied. The complaint doesn’t identify the owner of the sound recordings, nor does it need to do so.

The complaint doesn’t provide any detail about how the compositions were performed. Were the works performed by a live band? Did the Calvin Theater play a CD or stream the songs? Were the songs played over a radio? Which versions of the songs were played – the originals or cover recordings? The complaint doesn’t tell us, but it doesn’t matter – regardless of how the songs were played, the owners of the copyrights in the musical compositions are entitled to a public performance royalty. The music club should have paid that through a contract with ASCAP.  

What Is The Right of Public Performance?

One of the exclusive rights held by owners of musical compositions is the right to publicly perform a work. The Copyright Act defines public performance broadly – “to perform … at a place open to the public or at any place where a substantial number of persons outside of a normal circle of a family and its social acquaintances is gathered.” 

If a public venue plays popular music for its customers, it needs a public performance license. If it plays music and doesn’t have one, it’s a copyright infringer.

If you’re wondering about the public performance rights of owners of sound recordings – they don’t have one, except for digital audio transmissions. 17 U.S. Code § 106

What Are Publishing Companies?

The plaintiffs in this case are publishing companies. What’s that, you ask?

A lot of composers don’t have the time or inclination to deal with the music business. Instead, they assign ownership of their compositions to a publishing company to manage. There are many publishing companies, and the largest own publishing rights to thousands of songs. Two large publishers are the two plaintiffs in this case – Universal Music Publishing and Primary Wave. 

What Are Performing Rights Organizations?

The owners of compositions – whether publishing companies or the composers themselves – can’t track and police the thousands of public venues where their compositions may be performed. A complex system has evolved to deal with this –  they register their compositions with one of the PROs – ASCAP, BMI or SESAC. In turn, the PROs enter into blanket license agreements with the clubs and restaurants.

The clubs pay the PROs, the PROs pay the publishing companies, and the publishing companies pay the composers. This is all regulated by contracts – thousands and thousands of contracts.

It’s even more complicated than it sounds. For a deeper dive see Songtrust’s “Modern Guide to Music Publishing.“

The Publishing Companies Are Assignees of the Publishing Rights

The two plaintiffs in this case allege that they are owners of the six compositions that have been infringed. From this we know that the composers have transferred ownership of the compositions to these companies. The publishers must be either owners or exclusive licensees to bring a copyright infringement lawsuit.

Both publishers are members of ASCAP. In order to avoid infringing the compositions of these songs, Calvin Theater needed to enter into a blanket license agreement with ASCAP and not breach the agreement.

The Owner of the Club May be Personally Liable

The complaint alleges that Eric Suher is an owner, officer and director of Calvin Theater, that he controls, manages and operates the company that owns the club, and that he has the right and ability to supervise and control the public performance of musical compositions at the club.

This is important – in my experience many lawyers and business owners don’t realize that a corporation may not shield a business owner or manager from personal liability for copyright infringement. For details on why this may be the case, see Redigi – Did Ossenmacher Know He Was Risking Personal Liability?

If the publishers win their case Suher and Calvin Theater may be jointly and severally liable for copyright infringement. 

The Works Are Registered

To bring a claim of copyright infringement a work must be registered. This was uncertain until 2019, when the Supreme Court decided Fourth Estate Public Benefit Corp. v. Wall-Street.com, LLC. Before that some courts held that a pending registration was sufficient.

Here the publishers provided the registration numbers and dates for each registration. The compositions were initially registered in the early/mid-1970s. None of the copyrights have expired. In fact, the composers are still alive, so the copyrights will remain in effect for at least another 70 years. Even though the copyrights have been transferred, their duration continues to follow the lives of the composers. 

The Publishers Are Seeking Statutory Damages

The publishers don’t want actual damages (their lost profits) or the club’s profits attributable to the infringement – these are probably minimal. The publishers have asked for statutory damages of between $750 and $30,000 per infringing work.

Because the registrations preceded the infringements (which took place in 2022), they are entitled to seek statutory damages and, at the discretion of the judge, attorney’s fees.

However, the amount they are seeking is worth questioning – when an infringement is “willful” statutory damages may be as high as $150,000 per work infringed – in this case that would total $900,000. The publishers allege that ASCAP repeatedly told Calvin Theater that it was infringing and demanded that the Theater pay the contractual license fees. It’s not clear why the publishers are not seeking $150,000 per work based on what appears, at least for pleading purposes, to have been willful infringement.

That said, statutory damages are complicated. This table illustrates the options, depending on whether an infringement is “innocent,” “regular” or “willful”:

Conclusion

If the allegations are true this is a straightforward case. It illustrates the elements of a copyright case in the music industry, and how much trouble a public venue can get into by ignoring the requirement that it license rights from the performing rights organizations if it’s going to play popular music.

However, music publishers are not in business to force music venues into bankruptcy.  Most likely Calvin Theater’s  lawyers will tell their client that it should settle, and the publishers will accept reasonable terms.  I’ll keep an eye on the case and update this post if that happens. 

Update: The case was dismissed in December 2022. Very likely the dismissal was pursuant to a settlement. 

What You Need To Know About Goldsmith v. Warhol and Fair Use

by Lee Gesmer | Aug 22, 2022 | Copyright

The doctrine of fair use has been called, with some justification, the most troublesome in the whole law of copyright.  Justice Blackmun. Sony v. Universal (1984)

 

Fair use in America simply means the right to hire a lawyer.  Larry Lessig

 

Fair use is the great white whale of American copyright law. Enthralling, enigmatic, protean, it endlessly fascinates us even as it defeats our every attempt to subdue it.  Prof. Paul Goldstein

*********

The photo of Prince directly below was taken by Lynn Goldsmith in 1981. Andy Warhol used this photo to create an unauthorized series of sixteen silkscreens and drawings – the “Prince Series” – which appears below Goldsmith’s photo.

Conde Nast Cover

Goldsmith is a well-known rock-and-roll celebrity photographer. When Warhol passed away in 1987 the Prince Series became the property of the Warhol Foundation. Goldsmith was unaware of its existence until Condé Nast licensed one of the silkscreens for the cover of a Prince tribute magazine following Prince’s death in 2016. When Goldsmith learned that Warhol had copied her photo she sued the Warhol Foundation for copyright infringement.

Warhol’s defense in Goldsmith’s case is fair use – specifically the “transformative” branch of copyright fair use. This has its origin in Campbell v. Accuf-Rose, a 1994 case involving a parody of Roy Orbison’s song “Pretty Woman.” The Supreme Court held that a new work of art is “transformative” for purposes of copyright fair use if it “adds something new, with a further purpose or different character, altering the first with new expression, meaning or message.” 

This legal standard has proven to be subjective and inconsistent in its application. The Warhol case is a good example.

The District Court and Second Circuit Decisions in Warhol

A Southern District of New York district court judge agreed with Warhol’s defense that the Prince Series was “transformative.” The judge reasoned that while Goldsmith’s photo portrays Prince as “not a comfortable person” and a “vulnerable human being,” the Prince Series portrays the musician as an “iconic, larger-than-life figure.” Comparing the works side-by-side, the district court concluded that a reasonable observer would perceive that Warhol’s work has a “different character, a new expression, and employs new aesthetics with [distinct] creative and communicative results” when compared to the Goldsmith original.

The Second Circuit Court of Appeals disagreed. It held that to satisfy the “transformative” requirement the second work (the Warhol Series) must – 

. . . at a bare minimum, comprise something more than the imposition of another artist’s style on the primary work such that the secondary work remains both recognizably deriving from, and retaining the essential elements of, its source material. The judge must examine whether the secondary work’s use of its source material is in service of a fundamentally different and new artistic purpose and character, such that the secondary work stands apart from the raw material used to create it.

In the eyes of the Second Circuit Warhol’s silkscreens failed this test. Hence, they were not protected by fair use.

Interest in the case has been high since the Second Circuit issued its decision last year. It increased when the Supreme Court agreed to hear Warhol’s appeal, and has gone into overdrive as the case approaches oral argument on October 12, 2022. Warhol filed its appeal brief in early June. Goldsmith filed her opposition in early August. More than 30 amicus briefs have been filed. The Copyright Office and the Solicitor General have filed an amicus brief in support of Goldsmith, and the Solicitor General has asked for leave to participate at oral argument.

Google v. Oracle: Will It Matter to the Warhol Appeal?

An important consideration is how the Court’s 2021 ruling in Google v. Oracle may impact this case. Google is only the second time the Supreme Court has addressed fair use in depth. However, while the Court upheld Google’s fair use defense, the subject of that case was far from the traditional core of copyright – visual art, music and writings. Google involved fair use in the context of Google’s copying and reimplementation of Oracle’s Java API user interface. The Court found this to be fair use because it was socially beneficial – it allowed programmers familiar with the Java API to use their knowledge and experience to program Google’s Android operating system, rather than having to learn a new API. See Final Thoughts On Google v. Oracle.

Warhol argued that Google helped tip the scales in its favor, but the Second Circuit rejected this argument, stating that “a case that addresses fair use in such a novel and unusual context [as functional computer programs] is unlikely to work a dramatic change in the analysis of established principles as applied to a traditional area of copyrighted artistic expression.”

Will the Supreme Court affirm or reverse the Second Circuit? Setting aside Google (which is something of a one-off for copyright fair use), this is only the second time the Court will have addressed fair use since 1994 – will the Court expand fair use (by reversing the Second Circuit), contract it or tread lightly and leave it largely intact?

In pondering these questions it’s worth noting that changes in the Court’s make-up may be a significant factor in the outcome of this case.

Fair Use at the Supreme Court Without Justice Breyer

Until his retirement in June 2022 Justice Breyer had focused on intellectual property law more than any other member of the Court. He was viewed as the most liberal justice on IP issues, and he wrote the majority pro-fair use decision in Google. 

Given the current make-up of the Court post-Breyer, a little armchair kremlinology is in order.

Based on their dissent in Google it seems likely that Justices Thomas and Alito will vote to uphold the Second Circuit’s decision for Goldsmith. Under their view of fair use the most important factor is the effect of Warhol’s silkscreens on the market for Goldsmith’s photo. (Google, p. 1216). The Second Circuit found that Warhol’s silkscreens negatively impacted the market for Goldsmith’s original photo in a variety of ways, and Justices Thomas and Alito are likely to overweight this factor in concluding that Warhol’s silkscreens are not protected by fair use. 

The remaining justices on the new “conservative” wing of the Court – Justices Gorsuch, Kavanaugh and Barrett – favor “textualism,” the judicial philosophy that places primary weight on the normal meanings of a statute’s words, rather than public policy. It’s worth noting that the word “transformative” (indeed, the concept) appears nowhere in the Copyright Act, and is something of a judicial gloss on the statutory enumerated fair use factors. Based on a strict application of textualism these three justices may side with Justice Thomas’s view of fair use, in which case the Warhol Foundation will lose its bid to reverse the Second Circuit 5-4. If “swing conservative” Chief Justice Roberts joins the conservative wing, Warhol will lose at least 6-3.

My prediction: the Second Circuit’s ruling in favor of Lynn Goldsmith will be affirmed by at least a 5-4 vote.

Conclusion

Either way, affirm or reverse, will this case change fair use in the U.S.? We won’t know until the Supreme Court issues its decision, likely sometime in 2023. In the meantime, tune in to the oral argument in October and judge for yourself.

Update 5-18-23: I was correct in predicting that the Court would uphold the Second Circuit in this case. Here is the 7-2 decision – link

Is “Photo Embedding” Copyright Infringement? It Depends on the Court

by Lee Gesmer | Apr 27, 2022 | Copyright

If you own or manage a website you may be familiar with the process of “photo embedding” or “inline linking” an image or video on your site. Rather than hosting the image file on your own server you retrieve the image from another Internet site and embed the content as part of your webpage’s overall display. Users can’t tell the difference, but as a technical (and legal) matter you never “copy” the image to your server.

This practice is common, and has been performed millions of times on millions of sites. You have no legal concerns since you’re just channeling the image from another site. No worries, right?

Not so fast. This is actually a hotly disputed legal issue in the world of digital copyright.

For many years there actually were no worries. Way back in 2007 the Ninth Circuit created what has come to be called the “server test” in  Perfect 10, Inc. v. Amazon.com, Inc. This gave unauthorized embedding a legal green light so long as the image resided on another server. In copyright lingo the embedding site does not host any “material objects … in which a work is fixed … and from which the work can be perceived, reproduced, or otherwise communicated” and thus does not communicate a copy. 17 U.S.C. § 101.

However, in addition to the right of reproduction the Copyright Act gives the owner of a work the right of public display – copyright owners have the exclusive right to “transmit or otherwise communicate… a display of the work… to the public, by means of any device or process.” 17 U.S.C. § 101. The Ninth Circuit used similar reasoning to hold that an image is not displayed when the embedding site’s computer does not store the photographic images.

For many years the “server test” has been shaky but accepted law. A few courts around the country suggested that the Ninth Circuit’s holding that unauthorized embedding didn’t constitute public display was questionable, but there were no clear rulings on this issue.

However, in the recent decision in McGucken v. Newsweek LLC (S.D. N.Y. March 21, 2022) a federal district court in the southern district of New York (the “SDNY”) outright rejected this rule in a case in which Newsweek embedded a photographer’s Instagram post in an online news article without his permission. The court focused on the right of display and concluded that –  

After all, the Copyright Act defines “display” as “to show a copy of” a work, 17 U.S.C. § 101, and not “to make and then show a copy of the copyrighted work.” . . . The Ninth Circuit’s approach, under which no display is possible unless the alleged infringer has also stored a copy of the work on the infringer’s computer, would seem to make the display right merely a subset of the reproduction right. . . . The Copyright Act makes clear, however, that to “show a copy” is to display it. . . . Therefore, the Court finds that Defendant did in fact display Plaintiff’s Photograph when it embedded the Photograph in the Article.

In fact, this is the second time an SDNY court has reached this conclusion. In Goldman v. Breitbart News Network, LLC (2018) the district court held that an embedded tweet violated the photographer-plaintiff’s right of public display: “the plain language of the Copyright Act . . . provides no basis for a rule that allows the physical location or possession of an image to determine who may or may not have ‘displayed’ a work within the meaning of the Copyright Act.” (See my earlier post,  Is In-Line Linking Illegal Now?). 

However, in Goldman the Second Circuit denied interlocutory review and the case settled before any appeal. Hence, the issue didn’t reach the Second Circuit in that case. 

The split between the Ninth Circuit and the trial courts in the southern district of New York leaves website owners with a difficult decision – do they assume that Ninth Circuit law is controlling, and therefore continue to embed images? Clearly, for trial judges in SDNY the answer is “no” – these courts, at least, have made clear that they view this practice to be a violation of the right of public display. And, since a copyright plaintiff can often engage in “forum shopping” and file a copyright suit in that district the only safe conclusion is that this practice should be avoided nationwide, at least until one of the southern district cases reaches the Second Circuit on appeal and the split between the SDNY and the Ninth Circuit is resolved. If the Second Circuit creates a circuit split with the Ninth Circuit at the appellate level, the issue would be ripe for Supreme Court review.

There is another dimension to this issue that bears mentioning: the copyright issues are intertwined with complex terms, conditions and technical measures imposed by photo hosting services.

The poster child for this is Instagram. Instagram is the most popular source of images for embedded images, and it facilitates embedding by providing an embedding API. Late last year Instagram introduced a new feature allowing copyright owners to disable the embedding feature for photos they post. Before that, Instagram posters had to elect to use a private account to block unwanted embedding.

Instagram’s change to its embedding policy may reduce the likelihood of future copyright cases on this issue, but not by much. Many Instagram posters are unlikely to become aware of or implement this feature. Other social media and photo hosting sites, such as Twitter, Facebook and Flickr, have not followed Instagram’s policy. And, of course, there are millions of hobbyist and small-business sites that will not bother to block photo embedding, making it likely that this issue will continue to be the subject of litigation.  

Stay tuned. 

McGucken v. Newsweek LLC (S.D. N.Y. March 21, 2022)

Goldman v. Breitbart News Network, LLC 302 F.Supp.3d 585 (S.D.N.Y 2018)