Art in the ai of the beholder

HOW ARTIFICIAL INTELLIGENCE GENERATES ART—AND COPYRIGHT ISSUES

BY elizabeth ABREU

Copyright law is so fundamental to the progression of the arts that it is expressly enshrined in the Constitution.[1] This principle is based upon a belief that if people are granted a limited monopoly to something they created, they will be more incentivized to create and innovate. Today, advancements in the ability and accessibility of artificial intelligence, or AI, threaten the very sanctity of human creativity and authorship that copyright law is meant to protect.

AIs can easily copy a work into which a human author may have put substantial time and resources. Why would artists learn a craft if anyone can simply use AI to produce a nearly identical result with the same rights attached? Why would artists publish their works if AI can easily copy them? Why would companies hire employees to create if AI can reach the same results more cheaply?

The answers to these questions will greatly impact the future of the entertainment industry, from literary to audiovisual works. This article will walk through how AIs operate and AIs’ status in copyright law. Then, this article will discuss both the so-called input and output issues surrounding AI-generated art. Finally, this article will discuss specific implications of AI generators and copyright law in the entertainment industry.

How Generative AIs Work

AI models use inputted language to produce new works, such as images, audio, audiovisual works, text or code.[2] They use deep learning to produce this content.[3] Deep learning occurs when the AI is exposed to an abundance of data, observes this data and creates a statistical output based on it.[4] These AIs are called generative AIs because they use this statistical output to generate works similar to the data with which they were originally presented.[5] Specifically, generative AIs use a type of deep learning known as generative adversarial networks, or GANs.[6] A GAN has two networks: a generator network, which is responsible for creating the new data and improving it based on the second network, the discriminator, which is responsible for giving the generator feedback. This feedback allows the generator to emulate data as closely as possible to the prompts a user has given the AI.[7] An AI must be exposed to extremely large data sets to be effective because it can only create content based on the data it has been exposed to.[8]

AI collects data from different websites in a process called web scraping.[9] In this process, the data that is inputted into the AI is sourced from webpages brimming with potentially copyrighted works.[10] The AI is not knowledgeable as to what is and is not copyright-protected during this process, leaving the people who hold the rights to those protected works vulnerable to copying. A unique aspect of generative AIs is that their learning is unsupervised, so their developers cannot control where the AIs get the specific data.[11]

Judging by the reactions of artists to these AI generators, artists are very angry with the way AI has been scraping their art and replicating their styles.[12] This practice of web scraping could stifle people’s free access to copyrighted materials. Artists may have less incentive to create or publish their creations in accessible ways when AI can replicate their works with little effort.

The United States Copyright Office’s Current Stance on AI

Currently, AI generators are not able to obtain their own copyright.[13] In February 2022, the review board of the United States Copyright Office, or USCO, conclusively denied a work submitted by developer Stephen Thaler on behalf of his AI generator because it did not have sufficient human authorship.[14] This decision was rightly based on practicality, precedent and public policy. Thaler attempted to register the AI generated work with USCO three separate times and was denied each time for lack of human authorship.[15] Thaler has since filed a lawsuit against the USCO and the Director of the USCO, Shira Perlmutter, for violating the Copyright Act by denying the AI-generated work copyright protection and requesting that his application for copyright protection be reinstated.[16] In Thaler’s original complaint, he cited famed computer scientist Alan Turing’s test of artificial intelligence as the appropriate measure of whether a work created by AI should be protectable.[17] The test focuses on whether the product created by an AI is indistinguishable from a product created by humans rather than if the process that AI uses is indistinguishable from that of humans.[18]

Considering the product rather than the process does not comport with the rationale behind copyright law. Copyright law was written so that the granting of rights occurs as soon as the expression is fixed within its medium.[19] Therefore, one cannot overlook the process of how a work is fixed when determining who owns the copyright, regardless of the product seeming protectable.

Further, the implications of an AI holding its own copyright show why the USCO’s determination requiring human authorship for copyright protection should remain. For individual authors, copyright protection generally expires seventy years after the author’s death, depending on when it was created.[20] AI generators do not have a lifespan like human beings, so the limited monopoly that copyright law provides to a human author could become an absolute monopoly for an AI.

Copyright protection for works-made-for-hire expires either 95 years from the date of publication or 120 years from the day of creation if a corporation owns the copyright.[21] However, AI generators cannot be considered under the work-for-hire doctrine because this doctrine expressly requires that the work be authored by an employee or created under an express written agreement between parties.[22] AI is not capable of entering a legally binding employment contract as a party, so this doctrine is inapplicable.[23]

Another consideration of AI holding its own copyright is licensing. A corporation is an aggregate of human individuals, at least some of whom can act as an agent for the company and make an informed decision on whether the corporation would like to license a protected work.[24] AI generators do not have a client-facing aspect that could make an informed decision on whether to license the protected material. Consider a musical composition fully generated by an AI. For the purposes of granting permission to create a derivative work, who is the author? AIs do not have the capability to enter into licensing agreements to grant permission for others to make a derivative work based on a musical composition because they have no agency.[25] If there is no ability to license a copyright, a limited monopoly on a musical composition could become an absolute monopoly.

On March 16, 2023, the USCO released a statement of policy about works containing AI-generated material.[26] The USCO explicitly stated that the traditional elements of authorship are missing from works created by a human entering prompts into an AI system and receiving an artistic work in response.[27] In order for an AI-generated artistic work to gain copyright protection, the traditional elements of authorship must be determined and executed by the human being.[28] The USCO defines the traditional elements of authorship as “literary, artistic, or musical expression or elements of selection, arrangement, etc.”[29] The human author must also maintain creative control throughout the creative process, meaning that the human actively makes the creative decisions.[30]

When generative AI users type a prompt into the AI system, these human users do not assert any creative control over how the system interprets their prompts to generate artistic works.[31] It is more like the user is giving the AI an idea, and the AI is returning an expression based on that idea. The human contribution is not copyrightable in this instance because ideas are not copyrightable.[32]

For example, if a user tells the generative AI Midjourney to create an image in the style of Monet with rubber duckies in the water, the AI will determine the color scheme of the work, the placement of the rubber duckies, what other imagery is in the background, the size of the brushstrokes and the image’s focus, dimensions and viewpoint. Even if the prompts are extremely specific, the AI generator ultimately makes the creative decisions. Compare this to the level of effort required for computer programs like Adobe Photoshop, where the user must make all of the creative decisions to produce a work.

For works that have a combination of AI-generated material and human creative intervention, copyright protection could potentially be available depending on how much control the human author had when it came to the traditional elements of authorship.[33] In these instances, the human contributions will be the only copyrightable aspects of the work.[34] For example, there might be copyright protection if a human arranges AI-generated works in a creative fashion, or if an artist modifies an originally AI-generated work to such a degree that the author imputes traditional elements of authorship. This portion of the USCO’s statement was left intentionally vague so that the USCO could determine more detailed guidelines for these issues on a case-by-case basis.[35] Therefore, there is not a lot of guidance on how partially AI-generated works can receive protection.

The USCO states that copyright applications for works that consist partly of AI-generated material must specify that they contain AI-generated material and explain exactly which portions of the work were human-authored.[36] The USCO further states that works whose use of AI-generated content is more than de minimis should not be included in the application.[37] These guidelines show how the Copyright Office is attempting to protect human authors by not even considering to protect works that have a substantial use of AI-generated content.

The Copyright Office should implement a way to detect whether works are AI-generated. While USCO imposes a duty to disclose what aspects of work are AI-generated, the more realistic the AI becomes, the harder it will be to determine if AI generated a particular work.[38] The content produced by AI generators is becoming exponentially more realistic as the AIs learn through exposure to more data and more frequent use by people.[39] Additionally, the current tools to detect whether generative AIs are responsible for content creations are often not very accurate.[40] Better detection tools present a solution to a variety of issues that generative AIs present and can be used not only by the Copyright Office but also, for example, by consumers trying to determine if the news article they are reading has been written by AI or teachers attempting to determine if their students are cheating on essays. The companies that have generated these AIs should be generating solutions to issues they created.

Input Issue

Copyright infringement concerns arise from how AIs collect their data and where this source material comes from. Are the generative AIs committing massive copyright infringement by web scraping?

Getty Images seems to believe so, based on the complaint they recently filed in the District Court of Delaware against Stability AI.[41] Getty Images is a licensor that holds the copyright to millions of images across the internet.[42] The complaint alleges that Stability AI has removed or altered Getty’s copyright management information and provided false copyright management information on Getty’s database of over 12 million photos.[43] Getty’s database is especially useful for training generative AI because it pairs images with text descriptions.[44] Contextualizing these images allows the AI to learn associations more quickly.[45] Getty Images claims that Stability has taken from Getty Images’ database and used the database to get better at copying protected works and making the works it produces even more similar to protected sources.[46] Getty’s Terms of Use explicitly prohibit any data mining or copying of the images.[47]  Getty Images claims this prohibition did not prevent Stability AI from accessing and misusing Getty’s data.[48]

This practice is not uncommon. Midjourney’s CEO admitted that their generative AI uses over a hundred million copyrighted images without the rightsholders’ permission because permission would be practically difficult to obtain.[49] This creates a double-edged sword for artists putting their works in a public forum. As AIs continue to get more popular by scraping artists’ copyrighted works, artists may be disincentivized to put their work on the internet or share it publicly. For example, artists might limit their work’s accessibility online through password protection or two-factor authentication or by simply not posting their work online at all.[50] However, this would limit accessibility for the human audience and thereby stifle consumer engagement with the artists’ work. Consumer engagement is the lifeblood of artists, as art is often created to be seen or interpreted by others. Artists would be harmed economically because customers may not want to take the extra steps of creating an account or logging in to access the artwork, thus eliminating the possibility of people stumbling upon the art organically on the internet. These AIs could be committing mass copyright infringement by not asking permission before devouring artists’ data.

AI companies and researchers argue that the way AI collects data and alters it to generate a final product is transformative enough to fall under fair use and therefore deserves protection. But a transformation by a human being is distinguishable from a transformation by an AI. Just as the USCO holds that there is no copyright protection for a machine’s expression, there should be no copyright protection for a machine’s transformation, which is simply another form of expression.[51]

While the AI scrapes protected images, the AI makes creative decisions as to how the final product will be transformed from the originals it has scraped.[52] This includes decisions on what and how much to take from the original work, where to place it in the new work, what colors to change and how to resize it.[53] If the creativity of a work’s original authorship must come from a human, as the USCO says, then it follows that the creativity necessary for a work’s transformation should also come from a human.

Human beings have a conception of self and personal identity that exists within the past, present and future.[54] Humans transform works by drawing, either consciously or subconsciously, on their lived experiences, conception of self and dreams for the future. The purpose of a generative AI gathering data through web scraping on a technical level is to create an image that is as similar as possible to the images that are fed into it.[55]  By contrast, when human beings are faced with a work that they want to transform, they want to distinguish that work from the original, often commenting on or teaching about the original.[56]

To avoid this issue of infringing on others’ protected works, AI generators could create their own databases without protected works, but this would be an arduous task.[57] AI companies—which are, in part, motivated by the philosophy of making complex processes much simpler—are likely going to take the easy way out by utilizing what is already available, regardless of the consequences. Additionally, the amount of data the AI would be trained on would be much less, making the AI itself less accurate and thus damaging its commercial appeal. The commercial bottom line may be why AI companies are hesitant to change.

This practice of mimicking is extremely damaging to artists. Besides the lawsuit from Getty Images, three artists themselves have filed a class action lawsuit against Stability AI for copyright infringement.[58] The complaint makes frequent use of the word “betrayal” in reference to companies that have built AIs and the artists who use them.[59] While the law does not often consider emotional arguments, the plaintiffs are persuasive on this point.

For example, the artists argue in the complaint that AI has damaged their ability to develop personal styles.[60] Midjourney AI allows its users to create works based on the style of a variety of famous artists.[61] Although style cannot necessarily be copyrighted, the loss of the security in developing one’s own personal style has been significantly diminished.[62] Style is something that only humans can learn to develop and build over time; AI cannot create its own style because it must take note from prompts to make a new work derived from already-existing works.[63]  Human artists can visit museums, consume art online and take art history classes, among other things, for inspiration on how to craft their own style. But while human beings similarly take inspiration from other peoples’ styles, they also add different facets of their own experiences and personalities to their styles. Generative AIs do not have experiences and personalities to draw upon, so they must take only from others’ styles.

In the suit against Stability AI, the artists argue that their right of publicity is violated because their name is being used as a prompt to create an image in their style.[64] The complaint alleges that Stability AI had a feature that allowed the user to pick an artist so that their style could be emulated in the new work.[65] Stability allegedly attempted to draw more users to the site by advertising this to consumers and, therefore, was making a profit in connection to the name and style of the artist. The artists argue that people who utilize AI generators should not profit from other individuals’ uniquely curated styles.[66]

The Output Issue

Another issue stems from these AI generators’ output. Should the work made by the AI generator from the prompts of a human being be copyrightable without alteration? A human user could produce whole essays, novels, screenplays, song lyrics, etc. using ChatGPT. One could create whole visual universes using MidJourney, all from just a few words or phrases. If human beings are allowed to obtain copyrights for very detailed works into which they put minimal effort, the principle of propelling innovation will be stifled. Why would people take the time to create art when it can be generated in minutes from a few words? More daunting, why would companies employ human artists if generative AIs can create art without the complications of employment or time? While AIs give people the ability to partake in the artistic process more simply, that does not mean their products can or should be protected by copyright.

The creative arts are a different journey than other career paths. Those who set off on this path take a heightened risk that they will not succeed. This is so well-known that the colloquial term for someone pursuing a career in art is often a “starving artist.” Fewer people would be willing to take this risk due to artificial generators making their economic chances even slimmer. As a society, we should want to enrich our culture, not take away from it. Society is often propelled forward at the same time art is. Take the Renaissance for instance, where the best artists and scientists co-existed and were both highly innovative at that time.[67]

Art and science do not have to be at odds. AI creation is not inherently bad. It can be useful in helping humans become more productive in less time, and that is an ideal that pushes a capitalist society forward. But these interests must be balanced with real human interests. We should want to ensure that creative and economic opportunities for human beings are still at the forefront.

The standard recently set forth by the USCO leaves the open question of creating a standard for copyright protection that balances how much is human-authored and how much is AI-authored.[68] If the standard is too broad, it could lead to actual artists not receiving the protection they deserve, overshadowing the philosophical underpinnings of copyright law. If the standard is too narrow, it could have greater repercussions within the entertainment industry.

The answer to this question lies in the sanctity of art. Oxford Dictionary defines art as “the expression or application of human creative skill and imagination, typically in a visual form such as painting or sculpture, producing works to be appreciated primarily for their beauty or emotional power.”[69] How could an AI generator emulate these emotions through data sets? The idea behind entertainment is telling stories in creative ways. An AI has no experience, which is the crux of creativity in the entertainment industry. Songs are three- to five-minute stories. TV shows and movies are hours-long stories.  Even sports are rooted in tradition and the continuing story of the game that is being played. The art of storytelling has been passed down through generations, and it would be terrifying if this were the last line of human storytellers with genuine empathy and emotion, unrivaled by generative AI.

The USCO has thus far embraced and respected the sanctity of art because they have maintained that the traditional elements of authorship must come from a human being.[70] While AI can be used as a tool for fine-tuning, the essence of the expression should come from its human creator.[71] There has been a strange push towards romanticizing AI and giving AI human characteristics.[72] AI feels no purpose when it creates something, only that it is being told a task and is complying. Conversely, when human beings create, they feel a purpose behind the work.

For visual works, the human author should be the one physically arranging aspects of the work, placing objects or people into different aspects of the frame, editing color and color schemes and choosing the specific medium the digital work is supposed to mimic through hands-on editing.[73] For audio works, the human author should be the one moving around the notes or changing which instruments are featured and when.[74] The AI can then be used to clean up the work by smoothing things out or doing light editing, such as making lines straighter, ensuring a specific angle between lines or correcting other small mistakes.

An example of AI being used as a tool, rather than creating the traditional elements of authorship, is in the music industry. Here, AI has been used for years to generate pieces of musical compositions, which are then incorporated into a larger work by a human composer.[75] AI has also been used to generate rights-free music for content creators. In this way, AI-generated work’s place in the public domain has actually been beneficial to creatives.[76] Furthermore, AI has been used for around five years in mixing and mastering music.[77] In mixing, AI has been used to raise and lower the volume of different instruments or vocals at different points throughout the song.[78] In mastering, AI has assisted in smoothing out an album’s music tracks and in making sure that the quality of sound is consistent across the album.[79] These are perfect examples of how AI can be used to fine-tune creative works without commandeering creative control. 

Not only does the issue of AI generators affect those who create, but it also affects those who consume. In the internet age, where humans are already flooded with content daily, why would society want art created by something that cannot experience? In the highly monetized entertainment industry, AI-generated art may crowd out the starving artists. Instead, generative AI should be used as a tool to fine-tune a drafted work instead of maintaining creative control.

Specific Practical Issues in Entertainment

If works that are mainly generated by AI were able to gain copyright, there would be massive practical implications within the entertainment world. The entertainment world has been at the forefront of experiencing some of these issues, indicating what is to come.

Comic Books

A comic book whose illustrations were made using Midjourney was originally allowed copyright protection, but the protection for the illustrations was subsequently revoked.[80] The comic book’s arrangement of the images was copyrightable because it had an element of human creative control.[81] The artist who was attempting to register the work was involved in writing the prompts that generated the art and editing the images afterward, although it was unclear how much was edited.[82] The copyright office had an issue with how “random” and “unpredictable” the outputs of the AI were, meaning that the person typing in the prompts did not have sufficient creative control over the outcome to gain authorship.[83]

Music

The music industry has already seen a commercial failure with AI artist FN Meka. FN Meka was an AI persona voiced by a variety of real human beings. The AI artist was signed to Capitol Records for two weeks before controversy led to it being dropped.  The AI artist rapped about police brutality, and a photo later surfaced of its avatar getting hit by police.  This led to backlash over a computer-generated AI using the lived experience of Black people for profit.  This ties back to the sanctity of art and the discomfort experienced by consumers when interacting with art that is not based on lived human experiences.

TV/Film

Warner Brothers has already hired an AI company to determine which movies to reboot.[84] Is the next step using AI to actually make reboots? In the future, there could be screenplays adapted by programs like ChatGPT. Studios would be able to cut costs by eliminating human writers and replacing their work with that of a generative AI. If works partially created by generative AIs are not protected, the likelihood of this decreases, as studios, producers and distributors would not want to create a work that may be subject to the public domain immediately. The way that talent guilds function in the industry could also be altered by a surge of generative AIs producing content, as there could be less bargaining power on the side of the actual human beings due to lessened opportunity. As AIs become more sophisticated in crafting audiovisual works, they could alter the TV and film industries because of the way that these industries pull in a variety of different artistic mediums to create works.

An additional concerning example of AI’s impact on the industry comes from documentary editing or news reporting.[85]  Skilled operators could use AI generators to mimic participants’ voices and edit reality for the purpose of continuity.[86] While there are ethical rules surrounding news, there is no current code of ethics among documentarians.[87] This essentially means that everyone from documentary filmmakers to users on social media could utilize AI to slightly alter what interview subjects are saying to fit into the storyline the producers or creators are attempting to follow. Regarding concerns about altering what a person says or does, there should be ethical guidelines that forbid people from using AI to edit raw footage and alter what people say or do. Entertainment attorneys would be concerned if a standard emerged for contracts to contain provisions allowing others to use AI to create realistic representations where clients’ likenesses do or say things that the clients themselves did not do or say. It is a slippery slope to signing away a reputation.

Written Works

The Writer’s Guild of America, or WGA, has recently announced that it may begin accepting works written with the help of ChatGPT.[88] The WGA has outlined protections so that AI work cannot be considered source material or literary material for writers’ credit purposes.[89] This means that guild authors could take AI work, edit it and still receive union protection.[90] How much editing is necessary for protection? The USCO may require more editing by a human being for a work to be protected than the Guild does. Regardless of the credit protections for authors, the problem remains that ChatGPT could create a good amount of work product, with the traditional elements of authorship, and a writer could edit it and submit it. It may make it more difficult for unestablished authors to break into the industry, as studios would most likely just rely on the writers they already have to edit the material without looking for new sources for creative content.

Clearance

AI is going to make clearance exponentially more difficult for lawyers.[91] With how fast AI technology and knowledge are advancing, it may be difficult for lawyers to distinguish between AI-generated works and non-AI-generated works. If a work is AI-generated and a clearance review is required before the input issue is solved, how would the lawyer know where the AI took the art or literary work from or how substantially similar it is to art or literary works currently protected? The ability of AIs to generate audio works with a real individual’s voice or visual works with real people’s likenesses may lead to right of publicity and defamation issues.

Contract Law Remedy

If copyright law fails, can society prevent the plundering of its arts by AI through contract law? These AI services could help to solve the issue of their patrons exploiting any infringing works commercially by stating in their terms of service that users do not own their creations and cannot use them commercially. This is the case for most users of MidJourney, who must agree to terms of service granting to MidJourney a perpetual, royalty-free, irrevocable license to the images they create.[92] However, MidJourney users who choose to pay an additional subscription fee do own the works they create.[93] The same goes for any corporations using MidJourney that make over one million dollars annually.[94]  This may just lead companies to develop their own in-house generative AIs so that they can commercially exploit works without needing to pay additional subscription fees. There are also a lot of contract enforceability issues surrounding terms of service and assent, so this may not be the best approach.[95] 

Generating Solutions Instead of Issues

To adhere to the philosophy of copyright law and to protect the sanctity of art and artists, the law should create guidelines surrounding AI and copyright. By requiring that a partially AI-generated work has substantial elements of human authorship to qualify for copyright protection, we’re at a good starting point. AI will continue to generate such issues, so it is up to the law to generate equitable solutions in the future.

about the author

Prior to law school, Elizabeth Abreu (’24) attended the University of Michigan, where she worked on producing a film and for an organization that brought music to campus. This past summer, Elizabeth worked at PrizePicks, a daily fantasy sports company. Currently, Liz is a legal intern at Donaldson Callif Perez, a boutique entertainment law firm.

[1] U.S.Const. art. I, § 8, cl. 8.

[2] Decoding the Magic of Generative AI and How it Works, Tech Funnel (Dec. 22, 2022), https://www.techfunnel.com/information-technology/generative-ai/.

[3] Id.

[4] Ed Burns and Kate Bush, Deep Learning, Tech Target (last visited Mar. 26, 2023), https://www.techtarget.com/searchenterpriseai/definition/deep-learning-deep-neural-network.

[5] Id.

[6] Nick Routley, What is generative AI? An AI explains, World Economic Forum (Feb. 6, 2023), https://www.weforum.org/agenda/2023/02/generative-ai-explain-algorithms-work/.

[7] Id.

[8] Id.

[9] Sabrina Oritz, What is generative AI and why is it so popular? Here is everything you need to know, ZDNET (Feb. 15, 2023), https://www.zdnet.com/article/what-is-generative-ai-and-why-is-it-so-popular-heres-everything-you-need-to-know/.

[10] Id.

[11] Joe McKendrick and Andy Thurai, AI Isn’t Ready to Make Unsupervised Decisions, Harvard Business Review, (Sept. 15, 2022), https://hbr.org/2022/09/ai-isnt-ready-to-make-unsupervised-decisions.

[12] Complaint, Andersen et al v. Stability AI Ltd. et al, No 3:23-cv-00201-WHO (N.D. Cal. Filed Jan 13, 2-23).

[13] Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16190, 16194 (Mar. 16, 2023) (to be codified at 37 C.F.R. pt. 202).

[14] U.S. Copyright Office Review Board, Opinion Letter In re Second Request for Reconsideration for Refusal to Register A Recent Entrance to Paradise (Correspondence ID 1-3ZPC6C3; SR # 1-7100387071) (Feb. 14, 2022).

[15] Complaint, Thaler v. Perlmutter et. al, No 1:22-cv-01564 (D.D.C Filed Jun 02, 2022).

[16] Id.

[17] Id.

[18] Id.

[19] 17 U.S.C.S § 102. Subject matter of copyright: In general.

[20] How Long Does Copyright Protection Last? Copyright.gov, (last visited Mar. 26, 2023), https://www.copyright.gov/help/faq/faq-duration.html#:~:text=To%20determine%20the%20length%20of,and%20Circular%201%2C%20Copyright%20Basics.

[21] Id.

[22] Works Made for Hire. United States Copyright Office. (last visited Mar. 26, 2023), https://www.copyright.gov/circs/circ30.pdf.

[23]  Software Solutions Partners Ltd. v. H.M. Customs & Excise, [2007] EWHC 971 [67] (Admin) (noting that “on current authority,” “automated systems” could not enter contracts because “only a person with a ‘mind’ can be an agent in law”).

[24] Julia Kagan. Corporate Agent. Investopedia. (Jan 31, 2021), https://www.investopedia.com/terms/c/corporate-agent.asp#:~:text=Corporate%20agents%20allow%20companies%20to,will%20to%20make%20all%20decisions.

[25] Software Solutions Partners Ltd. v. H.M. Customs & Excise, [2007] EWHC 971 [67] (Admin) (noting that “on current authority,” “automated systems” could not enter contracts because “only a person with a ‘mind’ can be an agent in law”).

[26] Copyright Registration Guidance, 88 Fed. Reg. at 16190-16192.

[27] Id. at 16192.

[28] Id.

[29] Id.

[30] Id.

[31] Id. at 16190-3.

[32] 17 U.S.C.S § 102.

[33] Copyright Registration Guidance, 88 Fed. Reg. at 16192-3.

[34] Id.

[35] Id.

[36] Id. at 16913.

[37] Id.

[38] Id.

[39] See Burns and Bush, Deep Learning.

[40] Melissa Heikkilä, How to spot AI-generated text, MIT Technology Review, (Dec. 19, 2022), https://www.technologyreview.com/2022/12/19/1065596/how-to-spot-ai-generated-text/.

[41] Complaint, Getty Images (US), Inc. v. Stability AI, Inc. No 1:23-cv-00135 (D. Del Filed. Feb. 3, 2023).

[42] Id. at 2-3.

[43] Id.

[44] Id. at 3.

[45] Burns and Bush. Deep Learning.

[46] Getty Images, Inc. v. Stability AI Inc. . No 1:23-cv-00135.

[47] Id. at 13.

[48] Id. at 3-31.

[49] Matt Growcot, Midjourney Founder Admits to Using a “Hundred Million” Images Without Consent, PetaPixel, (Dec. 21, 2022), https://petapixel.com/2022/12/21/midjourny-founder-admits-to-using-a-hundred-million-images-without-consent/.

[50]Kevin Townsend, Web Scraping – Is it Legal and Can It Be Prevented? Security Week. (Nov. 7, 2022). https://www.securityweek.com/web-scraping-it-legal-and-can-it-be-prevented/.

[51]  See Copyright Registration Guidance at 16192-3.

[52] Decoding the Magic of Generative AI and How it Works; Ed Burns and Kate Bush, Deep Learning; Nick Routley, What is generative AI? An AI explains

[53] Id.

[54] Mark Turner, The Scope of Human Thought, On the Human, (Aug. 17, 2009), https://nationalhumanitiescenter.org/on-the-human/2009/08/the-scope-of-human-thought/.

[55] Id.

[56] Id.

[57] Id.

[58] Complaint, Andersen et al v. Stability AI Ltd. et al, No 3:23-cv-00201.

[59] Id.

[60] Id.

[61] MidJourney Style and Key Words Reference, (last visited Apr. 15, 2023), https://github.com/willwulfken/MidJourney-Styles-and-Keywords-Reference/blob/main/Pages/MJ_V3/Style_Pages/Just_The_Style/Artists.md

[62] See Complaint, Andersen et al v. Stability AI Ltd. No 3:23-cv-00201.

[63]See Decoding the Magic of Generative AI and How it Works; Ed Burns and Kate Bush, Deep Learning; Nick Routley, What is generative AI? An AI explains.

[64] Complaint, Andersen et al v. Stability AI Ltd. No 3:23-cv-00201.

[65] Id.

[66] Id.

[67] History.com Editors, Renaissance, History (Apr. 20, 2023), https://www.history.com/topics/renaissance/renaissance.

[68] Copyright Registration Guidance at 16192-4.

[69]Oxford University Press, Art, Encyclopedia.com, (Jun. 11, 2018). https://www.encyclopedia.com/literature-and-arts/art-and-architecture/art-general/art-history.

[70] Copyright Registration Guidance at 16192-4.

[71] See Id.

[72] Carys Craig & Ian Kerr, The Death of the AI Author, 52 Ottawa L. Rev. 31. (2021).

[73] See Copyright Registration Guidance at 16192-4.

[74] Id.

[75] Joe, The Rise of AI In Music – No Worries?,Thomann Blog, (Jan. 20, 2023), https://www.thomann.de/blog/en/the-rise-of-ai-in-music-2/.; Arek Skuza, The Increasingly Influential Role of AI in the Music Industry, Arek Skuza Blog, (Aug. 26, 2022), https://arekskuza.com/the-innovation-blog/the-increasingly-influential-role-of-ai-in-the-music-industry/.

[76] Joe, The Rise of AI In Music – No Worries, Thomann Blog.

[77] Id.

[78] Id.

[79] Id.

[80] Riddhi Setty, AI Comic Art Dispute Leaves Copyright Protections Open-Ended, Bloomberg Law, (Feb. 24, 2023), https://www.bloomberglaw.com/product/blaw/bloomberglawnews/bloomberg-law-news/XB7BH1H0000000?bc=W1siU2VhcmNoICYgQnJvd3NlIiwiaHR0cHM6Ly93d3cuYmxvb21iZXJnbGF3LmNvbS9wcm9kdWN0L2JsYXcvc2VhcmNoL3Jlc3VsdHMvNWRjYTYyZDkzNGI2N2JjYjFmNzVhYTk0NzY0YTcwOWEiXV0--e4d078005f34696a31e9b2434e941e7aa7d4f5fa&criteria_id=5dca62d934b67bcb1f75aa94764a709a&search32=PLMZtqvpNTsa7T_t9n2_yg%3D%3Dd9yWnrWAAWpXswL8nzT3RIBvXPbTs5wWSwMBq2aO0qZdqvvQRuCNaCYZL7FWZJuY1BWOwvKqsUvV77Bm7jbvG3eIJ-nfZ_DUXbizMt_ymvH-vP4jjzAzbwlmBwnHsOno.

[81] Id.

[82] Id.

[83] Id.

[84] Joe Berkowitz, Warner Bros. signs with AI company to help choose which movies to reboot (yes, really), Fast Company, (Jan. 9, 2020). https://www.fastcompany.com/90450218/warner-bros-signs-with-ai-company-to-help-choose-which-movies-to-reboot-yes-really.

[85] Mia Galuppo and Katie Kilkenny, Inside the Documentary Cash Grab, The Hollywood Reporter, (Sept. 16, 2022), https://www.hollywoodreporter.com/movies/movie-features/documentary-streaming-age-filmmaker-debate-ethics-payments-1235221541/.

[86] Id.

[87] Documentary Standards and Practices are Missing, Center for Media & Social Impact, (last visited on Mar. 26, 2023), https://cmsimpact.org/report/the-state-of-journalism-on-the-documentary-filmmaking-scene/p/documentary-standards-and-practices-are-missing/.

[88] Gene Maddaus, WGA Would Allow artificial Intelligence in Scriptwriting, as Long as Writers Maintain Credit, Variety (Mar. 21, 2023), https://variety.com/2023/biz/news/writers-guild-artificial-intelligence-proposal-1235560927/.

[89] Id.

[90] Id.

[91] Suzi Morales, The AI Revolution is Upon Us, Whether or Not Copyright Laws Are Ready, Observer (Mar. 30, 2023), https://observer.com/2023/03/the-ai-revolution-is-upon-us-whether-or-not-our-copyright-laws-are-ready-for-it/

[92] Terms of Service, Midjourney, (Feb. 10, 2023). https://docs.midjourney.com/docs/terms-of-service.

[93] Id.

[94] Id.

[95] When Are Terms of Use Legally Binding? SPZ, (last visited on Mar. 26, 2023), https://www.spzlegal.com/data-privacy/terms-of-use-legally-binding/.