The Internet Archive has lost a significant legal battle after the US Court of Appeals upheld a ruling in Hachette v. Internet Archive, stating that its book digitization and lending practices violated copyright law. The case stemmed from the Archive’s National Emergency Library initiative during the pandemic, which allowed unrestricted digital lending of books, sparking backlash from publishers and authors. The court rejected the Archive’s fair use defense, although it acknowledged its nonprofit status. This ruling strengthens authors’ and publishers’ control over their works. But it immediately reminds me of how AI tools train and use data on the Internet, including books and more. If the nonprofit Internet Archive’s work is not fair use, how do the paid AI tools use this data? 

Despite numerous AI copyright lawsuits, text-based data from news outlets usually doesn’t result in harsh rulings against AI tools, often ending in partnerships with major players.

You might think it’s different and argue that the Internet Archive directly uses books, but even though AI tools rely on all the data they have to generate your essay, you can still get specific excerpts or more detailed responses from them if you use a well-crafted prompt.

The Hachette v. Internet Archive case highlights significant concerns about how AI models acquire training data, especially when it involves copyrighted materials like books. AI systems often rely on large datasets, including copyrighted texts, raising similar legal challenges regarding unlicensed use. If courts restrict the digitization and use of copyrighted works without permission, AI companies may need to secure licenses for the texts used in training, adding complexity and potential costs. This could limit access to diverse, high-quality datasets, ultimately affecting AI development and innovation.

Additionally, the case underlines the limitations of the fair use defense in the context of transformative use, which is often central to AI’s justification for using large-scale text data. If courts narrowly view what constitutes fair use, AI developers might face more restrictions on how they access and use copyrighted books. This tension between protecting authors’ rights and maintaining open access to knowledge could have far-reaching consequences for the future of AI training practices and the ethical use of data.

Need a deeper dive into the case? Here is everything you need to know about it.

Hachette v. Internet Archive explained

Hachette v. Internet Archive is a significant legal case that centers around copyright law and the limits of the “fair use” doctrine in the context of digital libraries. The case began in 2020, when several large publishing companies—Hachette, HarperCollins, Penguin Random House, and Wiley—sued the Internet Archive, a nonprofit organization dedicated to preserving digital copies of websites, books, and other media.

The case focused on the Archive’s practice of scanning books and lending them out online.

The story behind the Internet Archive lawsuit

The Open Library project, run by the Internet Archive, was set up to let people borrow books digitally. Here’s how it worked:

• The Internet Archive bought physical copies of books.
• They scanned these books into digital form.
• People could borrow a digital version, but only one person at a time could check out a book, just like borrowing a physical book from a regular library.

The Internet Archive thought this was legal because they only let one person borrow a book at a time. They called this system Controlled Digital Lending (CDL). The idea was to make digital lending work just like physical library lending.

When the COVID-19 pandemic hit in early 2020, many libraries had to close, making it hard for people to access books. To help, the Internet Archive launched the National Emergency Library (NEL) in March 2020. This program changed things:

• The NEL allowed multiple people to borrow the same digital copy of a book at the same time. This removed the one-person-at-a-time rule.
• The goal was to give more people access to books during the pandemic, especially students and researchers who were stuck at home.

While the NEL was meant to be temporary, it upset authors and publishers. They argued that letting many people borrow the same digital copy without permission was like stealing their work.

Publishers’ riot

In June 2020, the big publishers sued the Internet Archive. They claimed:

• The Internet Archive did not have permission to scan their books or lend them out digitally.
• By doing this, the Internet Archive was violating their copyright, which gives them the exclusive right to control how their books are copied and shared.
• The NEL’s approach, which let many people borrow digital copies at once, was especially harmful to their business and was essentially piracy.

The publishers argued that the Internet Archive’s actions hurt the market for their books. They said people were getting free digital versions instead of buying ebooks or borrowing from licensed libraries.

Internet Archive’s defense

The Internet Archive defended itself by claiming that its work was protected by fair use. Fair use allows limited use of copyrighted material without permission for purposes like education, research, and commentary. The Archive made these points:

• They were providing a transformative service by giving readers access to physical books in a new, digital form.
• They weren’t making a profit from this, as they’re a nonprofit organization with the mission of preserving knowledge and making it accessible.
• The NEL was a temporary response to the pandemic, and they were trying to help people who couldn’t access books during the crisis.

They also pointed to their Controlled Digital Lending system as a way to respect copyright laws. Under CDL, only one person could borrow a book at a time, just like in a physical library.

The court’s decisions

District Court Ruling (March 2023)

In March 2023, a federal court sided with the publishers. Judge John G. Koeltl ruled that the Internet Archive’s actions were not protected by fair use. He said:

• The Internet Archive’s digital lending was not transformative because they weren’t adding anything new to the books. They were simply copying them in digital form, which wasn’t enough to qualify for fair use.
• The court also found that the Archive’s lending hurt the market for both printed and digital versions of the books. By offering free digital copies, the Internet Archive was seen as competing with publishers’ ebook sales.
• The court concluded that the Archive had created derivative works, which means they made new versions of the books (digital copies) without permission.

Appeals Court Ruling (August 2023)

The Internet Archive appealed the decision to a higher court, the US Court of Appeals for the Second Circuit, hoping to overturn the ruling. However, the appeals court also ruled in favor of the publishers but made one important clarification:

• The court recognized that the Internet Archive is a nonprofit organization and not a commercial one. This distinction was important because commercial use can often weaken a fair use defense, but in this case, the court acknowledged that the Archive wasn’t motivated by profit.
• Despite that, the court still agreed that the Archive’s actions weren’t protected by fair use, even though it’s a nonprofit.

Bottom line

The Hachette v. Internet Archive case has shown that even nonprofits like the Internet Archive can’t freely digitize and lend books without violating copyright laws. This ruling could also affect how AI companies use copyrighted materials to train their systems. If nonprofits face such restrictions, AI tools might need to get licenses for the data they use. Even if they have already started to make some deals, I wonder, what about the first entries?

Featured image credit: Eray Eliaçık/Bing

The contentious topic of AI copyright lawsuits is gaining traction, with many advocating that it’s high time AI enterprises compensated for the vast amounts of freely sourced data that have bolstered their generative systems.

In a recent wave of legal disputes, a multitude of lawsuits seeking remuneration from AI entities has emerged across the United States and Europe. The litigants range from individual authors and artists to large media conglomerates, all voicing their objections to AI’s appropriation of their creations for generating substandard offshoots.

An impactful open letter from the Authors Guild, bearing over 8,500 signatures from prominent writers such as Margaret Atwood, Dan Brown, and Jodi Picoult, has called upon the creators of generative AI applications, including ChatGPT and Bard, to halt the unauthorized use of literary works and to provide due compensation. These authors demand reparation for the data “harvested” to nourish these AI systems, likening it to an unpaid feast.

Writers also fear the potential for generative AI to undermine their craft by inundating the market with automated content derived from their original works. This concern was highlighted recently when Amazon had to intervene to address the issue of AI-generated books crowding its bestseller charts.

Before the Authors Guild made its appeal, authors Mona Awad and Paul Tremblay initiated legal proceedings against OpenAI. They alleged copyright infringement on the grounds that ChatGPT’s accurate summaries of their books implied the AI had been trained on their copyrighted material. They are not alone in this battle; author and comedian Sarah Silverman has also filed a lawsuit against OpenAI and Meta, accusing them of unauthorized replication of her autobiography, “The Bedwetter.” However, the intricacies of generative AI’s functionality might complicate the legal validity of these claims.

It’s not just individuals who are entering the legal fray. In a landmark move, The New York Times positioned itself as the first major American news outlet to bring a lawsuit against OpenAI, challenging the use of copyrighted material in the training and development of AI.

AI copyright lawsuits: The reason behind

The burgeoning phenomenon of AI copyright lawsuits is emblematic of a growing resistance to the unchecked use of copyrighted content by AI companies. While platforms like ChatGPT have been developed using internet-sourced data, they have done so without explicit consent from the creators of that data. Notably, GPT-3’s training encompassed a plethora of sources, including Wikipedia and Reddit. This process may inadvertently incorporate segments of copyrighted materials, enabling these expansive language models to concisely summarize copyrighted works with a disconcerting level of accuracy.

The issue magnifies when considering the enigmatic nature of AI. The “black box” dilemma, where the inner workings of AI remain obscured, exacerbates fears that AI could become a scapegoat for shirking accountability in both decision-making and content generation.

The legal contention also arises from concerns that if AI corporations continue to commercialize these opaque systems, these AI models could emerge as the quintessential means to an end. The danger lies in a potential future where decisions are not entrusted to AI systems for their efficacy or accuracy but because they can circumvent the legal and ethical constraints that bind human actions.

Data sources and methods

AI development, particularly with generative AI models like those at the center of numerous lawsuits, the process of data collection is a crucial and contentious aspect. The methods and sources from which these AI systems derive their training data have significant legal and ethical implications, especially when it involves copyrighted material.

Generative AI models, such as GPT-3 or ChatGPT, are trained on vast datasets collected from various online sources. These sources often include public websites like Wikipedia and Reddit, but can also encompass more contentious repositories like shadow libraries or other platforms where copyrighted materials are readily available. The training involves not just simple data scraping but also complex processes to understand context, style, and content nuances.

The legal gray area

The legal ambiguity arises from the fact that while the data is publicly accessible, the usage rights are not always clear. For instance, content from a public forum may not explicitly prohibit its use for training AI, but neither does it grant permission. This gray area has led to numerous AI copyright lawsuits, where plaintiffs argue that their intellectual property rights have been violated by the inclusion of their work in AI training sets without consent or compensation.

How AI is violating human rights?

AI technologies, while revolutionary, are increasingly scrutinized for potential human rights violations, a concern accentuated in the context of AI copyright lawsuits.

Key issues include:

• AI’s capability for extensive data collection and surveillance can infringe on individual privacy rights.
• AI systems can perpetuate biases present in their training data, leading to discriminatory outcomes in various sectors, underlining concerns in ongoing AI copyright lawsuits.
• AI-driven content moderation may inadvertently suppress free speech, an issue that intersects with the intellectual property debates in AI copyright lawsuits.
• In legal settings, AI tools can influence decision-making, potentially impacting the fairness of trials and judicial processes.
• AI-driven automation poses challenges to workers’ rights due to job displacement and the need for workforce adaptation.
• AI’s uneven access and impact can exacerbate existing inequalities, a concern that parallels the equitable access and usage rights at the heart of AI copyright lawsuits.
• AI systems that manipulate user behavior raise questions about individual autonomy and consent.
• AI’s control over information dissemination can affect the public’s right to access diverse and unbiased information.

Artists’ case for copyrights against AI faces uphill battle

What are the lawsuits against AI?

The legal arena is currently teeming with AI copyright lawsuits, with several cases spotlighting the tension between generative AI enterprises and copyright norms. The litigants include a variety of companies ensnared in these high-stakes legal battles.

Google: Data collection lawsuit

Google is facing a class-action suit accusing the tech giant of personal information misuse and copyright infringement. Allegations detail that Google harvested data, including images from dating sites, Spotify playlists, TikTok videos, and literature used to refine Bard. Launched in July 2023, the claim suggests Google might be liable for damages upwards of $5 billion. Opting for anonymity, the plaintiffs represent a growing concern over privacy and proprietary rights.

This spate of AI copyright lawsuits lawsuits is not without precedent. The Author’s Guild’s 2015 case against Google set a significant legal benchmark. The Guild challenged Google’s digitization of millions of books, offering snippets online. The ruling favored Google, characterizing the use as transformative and non-competitive with the original market for the books.

OpenAI: Copyright issues

OpenAI has also entered the legal fray, with authors Paul Tremblay and Mona Awad alleging copyright infringement. Their attorney, Butterick, represents a broader cohort of authors whose works, they claim, have been replicated within OpenAI’s extensive training data, potentially numbering over 300,000 books. Filed in June 2023, the lawsuit demands an undisclosed sum in damages.

OpenAI and Microsoft: NYT lawsuit

Additionally, The New York Times has launched a lawsuit against both OpenAI and Microsoft. The December 2023 filing contends that OpenAI utilized millions of Times articles to train their language models, which now rival the publication in delivering reliable information. Moreover, the lawsuit asserts that OpenAI’s models not only echo the unique stylistic flair of the Times but also recite its content verbatim. The Times, marking a first for a major American news outlet, pursued discussions regarding the copyright issue earlier in the year, but to no avail, culminating in this landmark litigation.

Meta and OpenAI: The Silverman case

Comedian Sarah Silverman’s legal action against Meta and OpenAI brings to light allegations of copyright infringement, positing that both ChatGPT and Meta AI’s Large Language Model (Llama) were developed using unlawfully sourced data that included her work. The lawsuit points to “shadow libraries” like Library Genesis, Z-Library, and Bibliotek, notorious for torrent-based content sharing, which often occurs without legal authorization. Specifically, the case notes that Meta’s Llama was informed by a dataset known as the Pile, compiled by EleutherAI, which purportedly contains data from Bibliotek. This suit was initiated in July 2023.

GitHub, Microsoft, and OpenAI: The Copilot controversy

A collective AI copyright lawsuits lawsuit targets GitHub, Microsoft, and OpenAI concerning the Copilot tool. This AI-powered service autocompletes code snippets by learning from a programmer’s input. The plaintiffs argue that Copilot unlawfully regurgitates code from GitHub’s repositories, disregarding the licensing requirements, including proper attribution. Beyond copyright complaints, the suit also accuses GitHub of personal data mismanagement and fraud. Filed in November 2022, the case has seen repeated dismissal attempts by Microsoft and GitHub.

Stability AI, Midjourney, and DeviantArt: The artistic integrity dispute

January 2023 saw a lawsuit against AI image generator companies Stability AI, Midjourney, and DeviantArt. Plaintiffs claim that these platforms infringe upon copyrights by training on and generating derivatives of the plaintiffs’ works. Additionally, there’s contention over the ability of these tools to replicate the styles of specific artists. The presiding judge, William Orrick, expressed a preliminary intention to dismiss the complaint.

Stability AI: The Getty Images lawsuits

Getty Images’ dual lawsuits against Stability AI spotlight the unauthorized copying and processing of countless images and associated metadata that Getty holds rights to in the U.K. A subsequent lawsuit in the U.S. District Court for the District of Delaware echoes similar copyright and trademark violations. It also emphasizes the concern over “bizarre or grotesque” images generated with the Getty watermark, potentially tarnishing the esteemed image repository’s reputation. These legal moves were made in January 2023.

Key questions raised by these AI copyright lawsuits

The emergence of AI copyright lawsuits signals a shift in how we view digital creativity. These high-profile legal confrontations raise several key questions that could redefine the copyright law in relation to generative AI:

1. Licensing for AI training materials: Is there a necessity for licensing when AI models are trained on copyrighted content? Given that generative AI systems replicate the training materials during their learning phase, the legal debate hinges on whether this replication falls under fair use or requires formal licensing.
2. Copyright infringement and AI outputs: Do the results produced by generative AI infringe on the copyrights of the materials used in training? A key aspect for the courts to determine is whether the similarities between AI outputs and the training data are based on protected or non-protected content. Additionally, the question of who bears responsibility for any copyright infringement committed by an AI system is yet to be resolved.
3. Compliance with digital copyright laws: Are generative AI technologies in breach of laws that govern the alteration or removal of copyright management information? This issue is particularly relevant in the case against Stability AI, where AI-generated images included false copyright management information, like reproduced watermarks.
4. Right of publicity and AI: Does creating AI-generated works that mimic the style of a specific individual infringe on their right of publicity? This right, which differs across states, restricts the use of an individual’s likeness, name, image, voice, or signature for commercial purposes without consent.
5. Open source licenses and AI: How do open source licenses intersect with the training and distribution of AI-generated content? This is a central concern in the GitHub Copilot lawsuit, where plaintiffs argue that the failure to attribute the source material and to release Copilot as open source violates the terms of open source licensing.

As these AI copyright lawsuits progress and begin to offer answers, entities involved in the development and deployment of generative AI tools should be attentive to emerging guidelines at the nexus of AI and intellectual property. It may also be prudent for these companies to consider strategies for mitigating potential risks in this evolving legal terrain. AI copyright lawsuits highlight the need for clear policies on data usage and rights.

Featured image credit: Igor Omilaev/Unsplash

Imagine a future where the artists of tomorrow are not only human visionaries but also lines of code and algorithms. This future is upon us, and it has ignited a legal battle that seeks to balance the scales of artistic expression and copyright protection in an era where machines are birthing masterpieces. In the heart of this storm stands U.S. District Court Judge William H. Orrick, who recently delivered a decision that could chart the course for the evolving relationship between AI and art.

This landmark lawsuit, brought before the U.S. District Court, brings three artists into the spotlight: Sarah Anderson, Kelly McKernan, and Karla Ortiz. Their accusation is clear and direct: AI art generators using the powerful Stable Diffusion technology have violated their copyrights, often without asking permission, paying, or even mentioning them. The defendants in this high-stakes showdown are Stability AI, Midjourney, and the iconic social platform DeviantArt.

AI art lawsuit breakdown

The lawsuit revolves around a complex issue of artificial intelligence, copyright law, and the creative arts. Here is a detailed explanation of the lawsuit:

• Parties involved: The lawsuit involves three main parties:
  • The Plaintiffs (Artists): Three artists, namely Sarah Anderson, Kelly McKernan, and Karla Ortiz, initiated the legal action. They have accused the defendants of copyright infringement related to their artworks.
  • The Defendants (AI Art Generators): Two entities, Stability AI and Midjourney, are creators of AI art generators that utilize Stable Diffusion technology to transform text into images. A third defendant, DeviantArt, is a popular social network and image sharing service that introduced its own AI image generator, “DreamUp,” using Stable Diffusion technology.
• Core allegation: The central allegation made by the artists is that AI art generators, including those produced by Stability AI, Midjourney, and DeviantArt, have been infringing upon their copyrights. The artists argue that these AI systems often employ extensive datasets of human-created art to train and generate new artworks. Importantly, this is done without the consent, compensation, or even awareness of the original human artists.
• The motion to dismiss: In response to the artists’ allegations, the defendants filed a motion to dismiss the lawsuit, seeking to have the case thrown out on various legal grounds. The primary objective of the motion was to argue that the lawsuit should not proceed based on the artists’ claims.
• Key defects in the complaint: The lawsuit faced several challenges due to what the judge deemed to be significant defects in the artists’ complaint:
  • Lack of copyright registration: One crucial issue highlighted in the case was that two of the artists, Kelly McKernan and Karla Ortiz, had not registered copyrights for their artworks with the U.S. Copyright Office. This omission was considered a significant weakness in their claims of copyright infringement.
  • Limited copyright registrations: Sarah Anderson, the third artist involved, had only registered copyrights for a fraction of her extensive body of work. This limitation further diminished the strength of the artists’ claims against the AI art generators.
• The role of the LAION database:
  • A key point of contention in the lawsuit was the Large-scale Artificial Intelligence Open Network (LAION), an open-source database containing billions of images. LAION was created by computer scientist and machine learning researcher Christoph Schuhmann and his collaborators. All three AI art generator programs under scrutiny relied on LAION as a crucial training resource.
  • The complexity arises from the vast and diverse nature of the LAION database. Judge Orrick emphasized that not every image within LAION could be considered copyrighted material, and not every artwork generated by AI was derived from copyrighted sources. This made it challenging to establish copyright infringement in each specific instance.
• The Challenge of “Substantial Similarity”: Judge Orrick underscored the requirement for a “substantial similarity” between AI-generated art and the original human-created works to establish copyright infringement. This meant that it was crucial to demonstrate that the AI-generated art predominantly drew from copyrighted material and bore a substantial resemblance to the original works. Without this requisite evidence, copyright infringement claims were unlikely to be upheld.

While the lawsuit faced dismissal due to the noted defects, the artists were allowed to amend their claims and refile a more focused lawsuit, specifically citing instances of infringed copyrighted material. Notably, the judge allowed one count to proceed—a direct copyright infringement claim against Stability AI related to Sarah Anderson’s 16 copyrighted works.

In summary, this lawsuit highlights the intricate and evolving challenges at the intersection of AI-generated art and copyright law. It emphasizes the need for continual adaptation in legal frameworks to address the unique complexities presented by AI in the realm of creative arts. While the lawsuit faced partial dismissal, it underscores the unresolved issues and ongoing debates in the evolving landscape of AI-generated art and copyright jurisprudence.

Featured image credit: Tingey Injury Law Firm

In a recent twist that has stirred conversations across the spheres of technology, art, and law, United States District Court Judge Beryl A. Howell’s ruling has shone a spotlight on the intricacies of AI-generated artwork and its place in the world of copyright. Here is the full ruling:

This decision, which poses that AI-generated art cannot be copyrighted without “human authorship,” opens a Pandora’s box of discussions about intellectual property, creativity, and the evolving relationship between humans and technology.

Beyond the brushstroke: AI as co-creator

Beyond legalities, the ruling prompts us to reconsider the very essence of artistry. Is AI merely a tool, an extension of human creativity, or can it be recognized as a contributor in its own right? This question delves into the heart of what it means to create and collaborate in the modern age.

As AI systems continue to play a larger role in generating art, music, and other creative works, we find ourselves at the crossroads of innovation and tradition.

A landscape beyond art

The ripple effects of this decision extend far beyond the canvas. As AI-generated content permeates industries like entertainment and media, similar questions of authorship and ownership come to the forefront. The ruling’s implications may reach into contracts, credits, and the very recognition bestowed upon those who contribute to AI-created content. This expansion of the conversation underlines the vast influence that AI wields in shaping the future of creativity and expression.

Charting a new horizon

Judge Howell’s ruling sparks a journey into uncharted territory, where AI and human creativity intertwine. The ruling catalyze conversations about the broader implications of AI’s presence in our lives, from reshaping the way we create to redefining the value we place on human ingenuity. It’s a discourse that transcends courtrooms and studios, inviting us all to ponder the evolving narrative of creativity and what it means for the future of art and innovation.

In an age where algorithms and human minds converge to shape our cultural landscape, the conversation around AI and copyright beckons us to reimagine the lines between the artist and the machine. As we stand on the precipice of an era defined by collaboration between human ingenuity and artificial intelligence, the question of who holds the brush becomes more nuanced than ever.

Ultimately, Judge Howell’s ruling serves as a thought-provoking chapter in the ongoing story of creativity’s evolution, inviting us to engage in a dialogue that paints a vivid picture of the road ahead.

Featured image credit: Tingey Injury Law Firm/Unsplash