Tag: Artificial Intelligence

AI models have an almost insatiable appetite for content. To date, the vast majority of training content, whether books or blogs or songs or images, has been scraped from the web without authorization from the copyright owners. Many AI companies argue that the use of copyrighted content in this context is “fair use” and does not require a license or compensation. But copyright owners are crying foul, and lawsuits are being filed at a rapid pace, including leading lawsuits from the Authors Guild and most recently the New York Times. Although fair use may be a defense to copyright infringement, only courts can determine if unauthorized copying is justified after performing a complex, fact-based, multi-factor assessment.

Big technology companies occasionally acknowledge that copyright owners deserve to be compensated, and a few early licensing deals have been concluded. Sam Altman, CEO of OpenAI, has repeatedly positioned himself as a moderate on the issue, “We’re trying to work on new models where if an AI system is using your content, or if it’s using your style, you get paid for that.”

However, other AI stakeholders are arguing that copyright owners are not entitled to payments for the use of their works to train AI platforms, and moreover, that any such requirements will stifle innovation and push it offshore. Writing to the US Copyright Office, the well-known VC firm Andreessen Horowitz put it this way: “The bottom line is this, imposing the cost of actual or potential copyright liability on the creators of AI models will either kill or significantly hamper their development.”

However, it is more likely that litigation rather than licensing will kill or disincentivize AI innovation. Companies may be less likely to use or develop AI technology if they may be sued. Moreover, companies will spend tens of millions to defend against lawsuits during the years it will take courts to sort out the law.  However, companies can move forward and establish license agreements now, either directly or through third parties. In fact, setting these commercial precedents will be likely to influence regulators in the future in a way that benefits copyright owners and AI developers alike, in part by proving that the commercial licensing of content will not be fatal to AI innovation. 

For AI development to continue at a rapid pace, a layer of service providers will need to be created to enable transactions without imposing additional friction into the system. A new class of service providers will be launched that will license content from aggregation points (i.e. publishers, agents, etc.), process the data with metatags and tokens, and license the data for training of AI models. This will provide an important legal precedent for the generative-AI industry, while creating new revenue-streams for copyright owners.

Calliope Networks is one such service provider. Calliope Networks is founded by executives with experience in AI and copyright licensing. Calliope Networks’ strategy is to aggregate books from authors and publishers and then to process the books for licensing and ingestion by AI models. Although AI systems have benefitted from the use of free content, new academic research suggests that higher quality aggregated works are significantly more valuable for effective training of large language models (LLMs) which underly generative AI systems. As a recent research report from Microsoft put it, “High quality data can… improve the state-of-the-art of LLMs, while dramatically reducing the dataset size and training compute.” 

One of Calliope Networks’ goals is to facilitate the legitimate use of copyrighted content by generative AI systems, ensuring that as AI evolves, so too does the respect for intellectual property. Calliope Networks is developing a platform that can operate at a speed and scale that can be an enabler to the growth of generative AI, rather than a burden.

Companies like Calliope Networks aren’t limited to the processing and licensing of original content. Calliope Networks in particular is also setting its sights on advanced products designed to enhance copyrighted works in a licensed, monetized fashion. Co-founder and CTO Jim Golden explains, “Our vision transcends mere compliance. We aim to expand the creative landscape by enabling ancillary AI-generated products that can, for example, enrich the experience of reading a book, but do so in a manner that ensures that the authors and publishers benefit appropriately.”

Copyright owners need to act to protect their right to control and be compensated for the use of their works. There is no doubt that the development of AI may represent a perilous and frightening future for creators. Nonetheless, the situation today is the absolute worst-case scenario: works are being used to create competing works, and copyright owners are not being compensated. If copyright owners choose to ignore the situation or simply refuse to demand commercial engagement with AI companies, courts and regulators may begin to believe the cries from Silicon Valley that the only way to ensure continued AI innovation is to broaden the fair use doctrine. 

Licensing will not be a panacea.  It may not provide all of the protections that publishers and authors will want.  Licensing will not protect against the eventual creation of AI-generated books for example. But refusing to engage in the licensing of copyrighted works may only serve to cede the battlefield to the technology companies and perpetuate the worst-case scenario of unauthorized use of creative works without compensation for years to come.

About the Author: Dave Davis is the CEO of Calliope Networks.  He previously served as Chief Commercial Officer of the Motion Picture Licensing Corporation and was an executive at Twentieth Century Fox, Paramount Pictures, NBCUniversal, and the Motion Picture Association.  Dave has a BA from Wesleyan University and a JD from the University of Michigan School of Law.  He can be contacted at Dave@calliopenetworks.ai

If you aren’t already a member of the Copyright Alliance, you can join today by completing our Individual Creator Members membership form! Members gain access to monthly newsletters, educational webinars, and so much more — all for free!

Fair use and AI has been a popular topic recently given the flurry of developments to generative AI systems. As discussed in the first installation of this blog, it is impractical to make broad predictions about how courts might rule on fair use and AI. However, after a high-level analysis of the four fair use factors, it is helpful to look at a few influential fair use cases to determine how they may be useful to guide us in determining how courts might analyze the four fair use factors when copyrighted works are used as training materials for AI. We picked the cases most frequently referenced when discussing fair use of training materials for AI.

Authors Guild v. Google

This is the case most AI developers reference when they say AI use of copyrighted training materials qualifies as a fair use. It is important to revisit the court’s analysis, because the principles from this case as applied to the typical AI-training materials fact pattern, particularly on the court’s transformative fair use reasoning, do not appear to be so clear-cut in favor of a fair use finding.

While the Second Circuit held Google’s copying of books to create a searchable database in Authors Guild v. Google qualified as fair use, its decision was limited significantly to the specific facts of that case. Those facts included not only the actions Google took to secure the reproductions of the books but also the fact that that Google was using the books to create a searchable database that would provide information about them. Instead of using copyrighted works to create a new product that could usurp the market for the underlying work, the court found that Google used the books to shed light on information about the book, for example, pinpointing where and how many times the word “whale” is used in Moby Dick. The court  noted that Google also took significant steps to secure the copies of books it used in its database, such as only showing “snippets” of works to highlight a search term and implementing anti-hacking measures. Due to these security measures, the court concluded that there was little risk that Google’s actions could serve as a substitute for the copied works.

Moreover, the court found that Google’s reproduction of copyrighted works did not create significant market harm for copyright owners. The Second Circuit held that, although snippet view would surely cause some decrease in sales, demand satisfied by snippet view would be for the work’s factual elements, not its creative, protected elements.

Unlike in Authors Guild v. Google, the generative AI training does not provide factual information about the copyrighted works. Instead, most generative AI reproduce and draw on the expressive elements from the copyrighted works as part of a process that results in works that would often act as market substitutes for the training materials—to say nothing of the harm caused to copyright owners who already offer licenses for AI training. A market exists for copyright owners to license their works for use in AI training datasets, and a court granting a fair use exception would destroy that market.

Texaco v. American Geophysical Union

Texaco v. American Geophysical Union may be a particularly important case in considering AI as an emerging technology. In Texaco, a commercial research company created an internal policy of photocopying and disseminating journal articles to hundreds of scientists. On appeal, the Second Circuit held that considering the existing licensing market for photocopying, the company’s practice harmed the publisher’s right to derive value from its copyrighted work. Although the licensing market in Texaco was relatively new and still developing, the court held that its existence weighed against fair use because the researchers had an alternative to infringement by simply licensing their photocopies. Although reproduction of the work was internal to the company and article copies were not distributed to the public, the Second Circuit focused on the fact that the Copyright Act grants the exclusive right of reproduction to the copyright holder. Texaco affirms the importance of this right, standing alone, even when the right to distribution has not been implicated. The research company also argued that its use was transformative because photocopying converted articles into a “useful format,” and because the copying aided in scientific research. The court disagreed, holding instead that a fair use exception cannot apply to a process, only to a work of authorship, and that the defendant could not “gain fair use insulation. . . simply because such copying is done by a company doing research.” The fact that the company’s research was done for commercial purposes further supported the denial of a fair use exemption.  

The general process of training generative AI on copyrighted materials shares several characteristics with the defendant’s photocopying in Texaco. The technological process of generative AI training seems to wholly reproduces a copyrighted work. Furthermore, like the burgeoning photocopying license market in Texaco, copyright owners offer AI training licenses. Just as the court ruled in the fair use case discussed above, Authors Guild v. Google, allowing a fair use exception for AI training would effectively destroy this licensing market.

Perfect10 v. Amazon.com

Perfect10 v. Amazon may become a battle ground for what constitutes a transformative means to an end rather than a non-transformative use. In this case, Perfect10 sued Google (and others) for its use of thumbnail versions of Perfect10’s copyrighted images on its search platform. The Ninth Circuit held that the defendants’ use was transformative because the thumbnails merely functioned as a “pointer,” providing social benefit by merely pointing to where a consumer could find the full images. The court’s decision was also influenced by the fact that because the image copies Google made were so small, they could not act as substituted for Perfect 10’s copyrighted works (and were also significantly smaller than the smallest resolution copies Perfect 10 licensed). It should be noted that Google’s use in this case did not merely repackage copyrighted works to recapture the artistic value they provided, the court found that Google instead created an entirely novel value by providing information about the works copied, which is not the case with AI generated art and the training of AI systems on copyrighted works.

Fox News v. TVEyes

Fox News v. TVEyes further demonstrates the tension between the first and fourth factors of the fair use analysis. In this case, defendant TVEyes offered a subscription service allowing consumers access to television programming to find, download, and share specific content like certain dialogue or other features of that programming. While the Second Circuit held that TVEyes’ service of increasing technological efficiency of viewing and sharing relevant clips was somewhat transformative, this transformative purpose could not outweigh the great harm to Fox by usurping its licensing market for clips. The court emphasized that transformativeness requires more than a “repackage” of a copyrighted work by “altering the [copyrighted work] with new expression, meaning or message.” Although TVEyes “modestly” transformed the work, the court held that the fourth factor outweighed the first because TVEyes “undercut[] Fox’s ability to profit from licensing searchable access to its copyrighted content to third parties.” The court correctly repositioned the importance of potential market harm and displacement of a copyright owner’s market within the fair use analysis.  

Generative AI trained on copyrighted works may not fare much better than the TVEyes program under a similar analysis. Many developers argue that training generative AI creates transformative value from copyrighted works. However, like TVEyes, these developers undercut copyright owner’s ability to license their works for AI development— a market that had already existed before the recent AI boom. The Second Circuit strongly emphasized the importance of the fourth factor in its analysis of TVEyes’ program, and the gravity of this factor, particularly in the light of existing markets must be given its proper weight in the fair use analysis of the use of copyrighted works in a generative AI training case.

Conclusion

The cases discussed in this blog may provide insight to how courts will rule in cases of generative AI training on copyrighted works. While future cases will be highly fact dependent, keeping in mind the principles of the fair use cases discussed in this blog may provide insight into how courts will rule in this area.

If you aren’t already a member of the Copyright Alliance, you can join today by completing our Individual Creator Members membership form! Members gain access to monthly newsletters, educational webinars, and so much more — all for free!

As lawsuits challenging the use of copyrighted works in generative artificial intelligence (AI) systems begin to snowball, there is one question at the forefront of the conversation: does the unauthorized use of copyrighted materials to train generative AI qualify for the fair use exception? As with all fair use determinations, the answer is that it depends; what is clear is that fair use does not magically excuse all AI use of copyrighted works as training materials. However, as courts begin to hear these cases, fair use and AI will likely remain at the forefront of copyright conversations.

Before further examining this question, it is important to explain that any fair use analysis is highly fact dependent. Anyone who makes overbroad statements that every use of copyrighted works for the purpose of training AI qualifies for the fair use exception, is patently wrong and has overlooked the nuances necessary to determining whether an infringing use may qualify for the fair use exception. The same can be said for those who make overbroad statements that every use of copyrighted works for the purpose of training AI is not a fair use. However, it is possible to make some educated generalizations about how a court might analyze the four fair use factors and to consider how existing case law may be applicable in situations where copyrighted works are used to train generative AI.  

First Fair Use Factor: The Character and Purpose of the AI Use

The first fair use factor, the purpose and character of the use, encompasses what will likely be a popular argument for proponents of the position that using copyrighted works to train AI qualifies for the fair use exception. The first factor takes into consideration whether the use is a commercial or nonprofit educational use, and whether the work “transforms” and adds something new to the copyrighted work. If a use is found to be non-commercial, that finding weighs toward a finding of fair use, while uses that are commercial weigh against a finding a fair use. However, the question of commerciality is just one part of the analysis of the first factor, and therefore is not dispositive.

In many cases, AI platforms trained on copyrighted works are used for commercial purposes. This is especially true of some of the most popular generative AI tools. Although some generative AI platforms may not initially appear to be commercial, platforms like Midjourney, Dal-E, and ChatGPT are now offering commercial subscriptions to their users (for example, OpenAI offers a subscription model for ChatGPT). Furthermore, even if an AI is not yet commercial, any plans AI developers make to enter the commercial market will also result in this factor weighing against a fair use finding. The bottom line regarding this part of the first fair use factor and AI is that courts are not likely to look kindly on an infringing use that reproduces copyrighted works to make money or otherwise financially benefit from their expressive, protectedelements. 

Instead, the crux of many AI developers’ fair use defenses will come down to the transformative purpose element of the first factor analysis, which has been an increasingly determinative (and amorphous) factor in fair use cases.

Many developers argue that training generative AI on copyrighted works is transformative because AIs scan works to identify and utilize “patterns inherent in human-generated media.” However, as will be discussed in the second part to this blog, transformativeness requires something more than reproduction for consumption, such as communication of information about the underlying work, or inclusion of information about the underlying work in a database. While findings of transformative uses have often disproportionately led to decisions that the use qualifies as a fair use, courts should still recognize that even if training AI is found to be a transformative use, that does not automatically mean that it qualifies as a fair use.

Second Fair Use Factor: Nature of the Copyrighted Work

Under the second fair use factor, courts analyze the nature of the copyrighted work; for example, whether the copyrighted work is factual or creative. Many generative AIs training on copyrighted works may run afoul of the second factor of this analysis because they are trained, at least in part, on highly creative works like visual art, music, or writings. While this factor is rarely dispositive, when the underlying work is creative, it weighs against a finding of fair use. To be clear, since fair use is a fact intensive analysis, there very well may be AI platforms that train on more factual works, and that may sway this factor towards a fair use exception.

Third Fair Use Factor: Amount of the Copyrighted Works Used

The third fair use factor considers the amount and substantiality of the portion used in relation to the copyrighted work as a whole. Courts have held that when an infringer copies the entirety of a copyrighted work, or the work’s creative “heart,” this factor almost always weighs against fair use, especially where multiple complete works are copied. Although in some cases copying of an entire work may be permissible because it is necessary to accomplish a transformative purpose, an infringer may take no more than necessary to achieve the transformative purpose.

The logistics of most AI generation and training involve a complete and total copying of multiple copyrighted works. In its recent complaint against Stability AI, for example, Getty Images describes that the reproduction of its high quality images, paired with detailed text descriptions has “been critical to successfully training the Stable Diffusion model to deliver relevant output in response to text prompts.” In other words, AI generators must copy as much as possible from expressive works, including the most expressive or crucially creative parts of the work, to achieve the purpose of training to generate quality output.  

In a statement to the USPTO, and relying on Authors Guild v. Google, Open AI argues that the amount of a copyrighted work copied is not the point of the third factor, but rather the amount of a copyrighted work made available to the public. Open AI admits that the use of entire works is “reasonably necessary” to create an accurate AI but argues that substantial copying should not matter when the copy is not made available to the public. This argument is completely unsupported by the Copyright Act, and if courts were to adopt this approach it would eviscerate the reproduction right by requiring a distribution to take place for a violation of the reproduction right to occur. Moreover, the fair use exception explicitly directs courts to look at the amount and substantiality of the copyrighted work that is used as opposed to a judicially created “public access” theory. Add to that the fact that while OpenAI and other developers say that copies are not made available to the public, it’s unclear whether or how repositories of works that are created without authorization for training purposes are safeguarded against further reproduction and distribution. Although the third factor is not dispositive and (like the other factors) is highly fact dependent, where works are produced in their entirety, this factor will likely weigh against a fair use exemption.

Fourth Fair Use Factor: The Impact on the Value and Market for the Copyrighted Work

The fourth factor of a fair use analysis weighs the effect the infringing use has on the potential market for or value of the copyrighted work. Courts have held this factor to weigh against a fair use finding when the infringing work acts as a market substitute for the copyrighted work, and sometimes even when an infringing use lays outside the markets a copyright owner currently occupies (so long as that market is one a copyright holder might reasonably enter).

There are strong arguments to be made that AI training on copyrighted works harms the market and value for those copyrighted works. Foremost is the fact that many developers do not compensate copyright owners for the works used to train generative AI, despite the fact that many copyright owners presently offer AI training licenses. This destroys copyright owners’ licensing markets. Generative AI systems trained on copyrighted works harms the market for copyright owners to license works for AI training datasets. Many artists and copyright owners offer licenses for their works to be included in AI training datasets. Getty Images for example offers licenses for AI developers to use its images in training datasets (licenses which Stability AI did not obtain). Other copyright owners like academic publishers and others also offer AI licenses. The continual development of these licensing markets points to the fact that copyright owners are part and parcel of the AI developing world and are working with or are open to working with AI developers to advance AI innovation and tools. Courts have recognized where such a viable market exists to help artists capture value from reproduction and distribution of their copyrighted works, potential licensing value from that market should be considered in a fair use analysis.

Conclusion

As discussed, fair use cases involving generative AI training on copyrighted works will be highly fact dependent. While some AI-related uses may qualify as fair use, unauthorized use of copyrighted material to train AI systems cannot be handwaved by a broad fair use exception that disregards the rights of creators. Neither the Copyright Act nor case law (as will be discussed in the second installment of this blog) would support such a broad fair use exception for AI. Without the factual nuances of a real application, it is difficult to say how courts may decide fair use and AI cases. However, while technological innovation will often test our understanding and application of fair use, the underlying principles of copyright law must not be cast aside in favor of in an unhinged race of technological advancement which may lead to harmful and irreversible consequences. AI should be responsibly and ethically developed, and developers must respect artists and copyright owners’ rights.

If you aren’t already a member of the Copyright Alliance, you can join today by completing our Individual Creator Members membership form! Members gain access to monthly newsletters, educational webinars, and so much more — all for free!

It seems that all over the news, Artificial Intelligence (AI) dominates the headlines. Copyright is no exception. Given its recent focus on the scope of copyright in an AI-assisted work and in an AI generated work from two-high profile registration cases, the U.S. Copyright Office recently published a Copyright Registration Guidance (Guide) to further explain its examination practices and policies when reviewing registration applications for works containing AI generated material.

The Office cautions that while it is tasked with reviewing the applications, it is ultimately up to the applicants “to disclose the inclusion of AI-generated content in a work submitted for registration and to provide a brief explanation of the human author’s contributions to the work.”

While creators and copyright owners should complete the registration application as truthfully and as accurately as they can, they can do so with the comfort of knowing that a mistake on the registration application will not ultimately invalidate the registration. Also, applicants must remember, that while registration provides many benefits, it is not a necessary requirement for copyright protection because copyright protections are automatic from the moment a copyrighted work is created (regardless of whether or when it is registered with the Office). So if a work is not registered or a registration is deemed to be invalid, it does not mean copyright laws suddenly don’t apply to the work.

Nevertheless, registration is crucial for creators and copyright owners to fully enjoy their rights under the Copyright Act, since registration enables the copyright owner to do things like bring a federal lawsuit and to recover statutory damages in federal court. Since the Office recently granted a copyright registration for portions of an AI-assisted work, it shows that creators can register works created in part using AI. But determining levels of human authorship versus AI generations can be tricky. Based on the Office’s Guide, here are the three takeaways to keep in mind when registering the copyright for an AI-assisted work.

1. Descriptions of Human Authorship on an Application Must Go Beyond Asserting That Creative Prompts Were Entered into an AI Machine

In the Guide, the Office stresses that an applicant must describe the human authorship in the creative elements of the work being registered in the “Author Created” section of the application, and disclaim AI-generated content in the work that is more than de minimis (more on that in takeaway number two).

This is important because a Copyright Office examiner will be analyzing this information and will follow the Office’s policy to refuse registration for a work generated by AI, or for the parts of a work generated by AI, on the basis that AI generated content do not have human authorship.

The Office doubled down on that policy, explaining in the Guide that when it considers the copyright in AI-assisted works it will mainly observe “whether the AI contributions [in the work] are the result of “mechanical reproduction” by the AI or is instead “an author’s ‘own original mental conception, to which [the author] gave visible form.”

The Office illustrates the difference between such “mechanical reproduction” by AI and human authorship in an example where a poem is generated by AI based on a prompt from a human, “write a poem about copyright law in the style of William Shakespeare.” In the example, the Office takes the position that the AI determines the results in the expressive elements of the generated poem like the rhyming pattern, the words in each line, and the structure of the text — not the human who entered the prompt. Later on in the Guide, the Office more explicitly states that entering a prompt (even if the prompt itself may be a creative, copyrightable expression) into an AI machine alone “does not mean that the material generated from a copyrightable prompt is itself copyrightable.” In other words, a registration applicant won’t be able to claim human authorship in AI generated materials or elements in the work by merely stating that they entered prompts into an AI machine.

Some applicants may have difficulty with the Office’s position, as there are arguments that entering certain types of prompts (either one, or many) into an AI machine may result in sufficient creative ideation and control which amounts to human authorship in the AI generated content. But based on the Guide and at least for now, it may be quite difficult for applicants to convince the Office otherwise for registration purposes.

The Office subsequently states that applicants must detail their creative choices made to the AI generated content, such as the creative selection, coordination, and arrangement of that content in the work being registered. For example, if an applicant incorporates AI-generated text into a larger textual work, the Office states that the applicant should “claim the portions of the textual work that is human-authored” (i.e., the creative elements not generated by an AI machine).

2. Applicants Must Disclaim AI-Generated Content in the Work that is More Than De Minimis

As noted previously, the Office also asks applicants to explain what parts of the work should be excluded from the registration application. In its normal registration practices, the Office does ask registration applicants to disclose such information, since the scope of copyright protection for a registered work does not extend to certain parts of a work, like previously published and public domain materials.

Because the Office takes the position that a registration does not cover materials resulting from non-human authorship (i.e., there is no copyright in AI generated content) the Office directs applicants to detail “AI-generated content that is more than de minimis to be excluded from the application.” The Office points applicants to its examination practices on the registration of unclaimable materials and definition of de minimis in its Compendium, which states:

“Creative authorship is deemed “de minimis” when a work does not contain the minimal degree of original, creative expression required to satisfy the originality test in copyright.”

In providing examples, the Office notes that materials like brief quotes and short phrases may fall into that de minimis category. So, for any AI generated materials in a work that falls above that threshold, the Office will look to applicants to describe those disclaimed materials so that the registration can reflect the scope of copyright in the AI-assisted work.

3. Works Containing AI Generated Materials Must Be Registered Using the Standard Application

One of the biggest takeaways from the Guide is that the Office is requiring applicants seeking to register the copyright in their AI-assisted works to only use the Standard Application form. In a footnote, the Office reasons that its other types of application forms currently don’t contain fields where applicants can disclaim unprotectable material, like AI-generated content. As an example, the Office notes that its Single Application “may only be used if ‘[a]ll of the content appearing in the work’ was ‘created by the same individual.’” Presumably, this also means that applicants cannot use the group registration options to register their works that have AI-generated elements. But as more creative communities utilize and develop AI into their works, the hope is that the Office will increase the flexibility of the registration system to accommodate for the registration needs of creative industries and creators who want to register the copyright in their AI-assisted works, beyond the Standard Application, such as those who frequently use the current group registration options.

Just Do Your Best!

If applicants are still unsure of how to fill out the application according to the guidelines the Office set forth, the Office also notes that applicants can simply provide a general statement that the work contains AI-generated material so that the registration examiner can follow up with the applicant. However, this can result in a prolonged examination time because correspondence between the applicant and the Office would lengthen the amount of time it takes for the Office to ultimately issue a copyright registration for the AI assisted work. Taking time to think through the human authorship expressions and AI generated materials will inevitably help an applicant to streamline the registration process. Applicants can also sigh in relief that even if they made a mistake on their application form, their registration won’t be invalidated in court. There are also opportunities to directly reach out the Office to correct pending applications and to file supplementary registration to correct granted registrations.

The Office continues to study the implications of AI in copyright law and will be holding a series of public listening sessions this spring. But with this Guide, they have set a few guideposts to how applicants could register the copyright in AI assisted works. Though authorship issues, especially in AI, can be confusing for applicants to think through, we hope some of the pointers above will help registration applicants along in the process.

If you aren’t already a member of the Copyright Alliance, you can join today by completing our Individual Creator Members membership form! Members gain access to monthly newsletters, educational webinars, and so much more — all for free!

The exponential development of Artificial Intelligence (AI) systems represents a profound achievement of the digital age that brings with it tremendous opportunities. In fact, many in the creative community are already using or plan to use AI for the creation of a wide range of works and are developing and pushing AI capabilities to explore new creative horizons. But as with many advances in technology, these new opportunities come with challenges and often raise difficult legal questions.

Artists, authors, and many other types of creators are increasingly concerned (and rightly so) about the tendency of some to ignore or discount issues relating to copyright in the AI context. Though the application of copyright law to AI may be tricky at times, it is essential that these issues not be ignored or given short shrift in the AI discussion just because AI is the shiny new toy. Rather than sweep important copyright issues under the rug for fear of slowing AI’s progress, policymakers, lawmakers, stakeholders, and the public must respect the rights of creators and copyright owners and recognize and appreciate the underlying goals and purposes of our copyright system.

The Relationship Between AI and Copyright Law

Well before the recent explosion in AI development, the creative community has been using AI technologies and innovating in the space. On the output side of AI, creators and copyright holders use or actively develop AI technologies as part of their larger creative process. For example, video game developers utilize AI systems and technologies to provide new and improved gaming experiences, such as when players interact with in-game, non-player characters. Television and filmmakers also often incorporate computer generated images (CGI) into their works that are created with the help of AI programs. In the music industry, artists and producers are employing AI tools for everything from beat creation to voice modulation.

On the input side of AI, the primary way the creative community drives AI innovation and development is by creating and disseminating copyrighted works that are used to train AI systems. At best, the creative community collaborates with and innovates as a part of the AI community by developing and improving AI technologies, licensing works for use as training material, or using AI as a tool to generate or make new works. At worst, copyrighted works are used by AI technologies without authorization or licenses—sometimes for the purpose of creating works that serve as direct market substitutes for the ingested works—undermining the rights of artists, creators, and copyright owners and their abilities to protect, license, and enforce their copyrights.

Sadly, it is the latter situation that is becoming more widespread in the AI world where the works of countless creators and copyright owners are being used without permission or compensation. This especially true and especially harmful with commercial AI uses.

Using Copyrighted Works as AI Inputs

An AI’s creative output is only good as the corpus of creativities it ingests. For example, if a human prompts an AI machine to generate an image to accurately imitate the work of a famous visual artist, like Jean-Michel Basquiat, the AI machine must analyze and copy the expressions that are unique to Basquiat’s works. But what is sometimes lost on AI system developers and users is that the underlying works, which the AI draws from, are more often than not created by a human creator. That human creator depends on the rights and protections granted to them by copyright law to commercialize and control their works, including the ability to license their works for use as AI input (or to stop others from using their works without authorization).

Creators and copyright owners also make contributions to the development of AI technologies by priming copyrighted works for optimal AI application and development through activities like semantic enrichment, metadata tagging, content normalization and data cleanup. Copyright owner-curated and prepared AI training data sets, databases, or collections of works also feature additional benefits like secured licensing and permissions from third parties, which reduce privacy and infringement risks for AI developers and users. It is copyright laws which incentivize and protect the investments creators and copyright owners make when creating and preparing these kinds of works for AI.

Text-and-data mining (TDM) is the process through which AI machines develop their unique algorithms and capabilities by analyzing, reproducing, and otherwise using valuable data and expressions often contained in copyrighted materials. TDM uses input materials to develop trends, algorithms, and methods that can generate output works. During TDM, an AI machine might very well be analyzing numbers, statistics, and other non-copyrightable pieces of information. But AI machines that generate output like images, videos, or songs, risk displacing or substituting for the very underlying copyrighted works that are used to train the AI. During TDM of copyrighted works, the AI machines are often culling the expressive, copyrightable value contained in the ingested works, as mentioned previously with the Basquiat example.

Many creators and copyright owners currently offer TDM licenses so their works can be used to train AI systems. Copyright law enables creators and copyright owners to create works that are fuel for AI development, and there must be respect and recognition of the laws and rights that protect their ability to license (or not license) their works. Particularly where AI machine outputs serve as replacements or substitutes in the markets for the ingested works, any artificial disturbance or heavy-handed approaches to manipulate the existing market for TDM licenses for copyrighted works, without being supported by evidence, results in creators and copyright owners subsidizing AI development. When works are used without authorization, licensing markets—the fundamental ability of a copyright owner to control and commercialize their works—are effectively destroyed.

No Broad Exceptions or Justifications Exist for AI Use of Copyrighted Works

The disturbing reality is that many AI companies do not license copyrighted works to train AI machines. Nor are they transparent about the sourcing of their training data sets. Some companies simply scrape existing copyrighted content from the internet including images, text, and software code to use as training data sets. Other companies engage in a practice called “data laundering” where they fund or use data sets created by academic or research institutions for initially noncommercial purposes to train commercial AI machines.

In the discussion surrounding the use of copyrighted works to train AI systems, some stakeholders wrongly justify these practices with the argument that the fair use exception would wholesale permit such methods or could justify broad copyright exceptions for AI use. That view is inaccurate— especially in a case where a TDM license is available, the use is commercial, or the resulting AI generated work harms the actual or potential market for the ingested work. Fair use is such a fact-specific exception that it is an unreliable basis to build any broad AI exceptions on or to make general claims that these AI practices are excused from blatant copyright infringement.

AI and Copyright Laws Around the World

For many years now, lawmakers and policymakers in a number of countries, including the United States, have been carefully examining the intersection of copyright law and AI and the implications of this rapidly evolving technology. Even as a global leader in AI technologies, the United States has not deemed it necessary to enact or recommend any new exceptions to copyright law for AI purposes. And for good reason: licensing to support AI application is robust and without contrary evidence there is potential for significant harm by prematurely upending creators and copyright owners’ copyrights.

The United States is not alone in its treatment of the AI licensing market and its relation to copyright law. Very few countries have considered AI regulations and policies with respect to copyright laws including, Hong Kong, South Korea, Australia, and Canada. Significantly, each country has declined to take action, postponed decision making as premature, or otherwise not taken action. In varying degrees, only the European Union, Japan, Singapore, and the United Kingdom have AI policies and regulations within their copyright laws.

One example of a country with problematic AI-copyright regulations is Singapore, which overbroadly permits unauthorized TDM of copyrighted works, including for pirated works, for any purpose with no ability for rightsholders to opt out or contract around the exception. Policies such as Singapore’s severely undermine the fundamental ability of creators and rightsholders to be compensated for the use of their copyrighted works, discouraging them from creating the works and depriving them of the fruits of their labor. Unfortunately, the United Kingdom is also considering following this troubling precedent, with a proposed exception for TDM of copyrighted works for noncommercial and commercial uses, with no ability for creators and copyright owners to contract around the exception. Needless to say, many in the U.K. creative community have decried the proposal, and we can only hope that the U.K. will reverse course to avoid undermining a critical part of their economy.

Key Principles and Takeaways for Examining AI and Copyright

As AI technologies continue to progress at a rapid pace and make incredible advancements, AI stakeholders, courts, policymakers, and the public should keep in mind several key principles when analyzing the intersection between AI and copyright.

• When formulating new AI laws and policies, it is essential that the rights of creators and copyright owners be respected.

• Long standing copyright laws and policies must not be cast aside in favor of new laws or policies obligating creators to essentially subsidize AI technologies.

• Education is paramount in the AI space. Those leading AI projects are aware of the legal implications of using copyrighted works input material, and those that arise from AI-generated output

Along these lines, we the Copyright Alliance recently published our position paper on AI and copyright law issues that outline the above key principles and also points out some other detailed positions of the copyright community on AI. It is critical to AI innovation that the creative contributions and impact of the creative community to AI are acknowledged and that the foundations of copyright law that made AI possible in the first place are preserved and respected.

If you aren’t already a member of the Copyright Alliance, you can join today by completing our Individual Creator Members membership form! Members gain access to monthly newsletters, educational webinars, and so much more — all for free!