News Post

Getting your Trinity Audio player ready...

By Benjamin P. Mayer

A Google search of “What AI thinks Americans from each state look like” can yield some interesting results. Numerous people have used artificial intelligence tools to generate output images of Americans from each state, and the images of what AI thinks about Alabama are pretty much what one would expect: stereotypical, offensive, hilarious and/or accurate, depending on how you feel about the state. We’ve all seen many of these images being used for monetary gain as well. These days, it is easy enough to copy and paste a cartoon-y AI output image of an Alabama yokel, slap a caption on it, and sell it online for profit.

Should you be required to pay royalties for using that image? If so, who owns the copyright on an image not “authored” by a human? With the use of AI on a meteoric rise and, seemingly, infiltrating countless aspects of our lives, the need for guidance in anticipation of regulation is a constant. The Congressional Research Service issued a Legal Sidebar report summarizing issues and opinions pertaining to this topic.[1]

But first, a bit of background on copyright law is warranted. The federal government provides copyright protection for original works of authorship to the authors of such works. This form of intellectual property allows artists, authors, and creators to publicize and distribute their creative works to the public without fear of interference by infringement and piracy. Copyright exists to incentivize creation of works that benefit the public. The holder of copyright protection over a work has the power to capitalize on the economic value of the work through authorization of certain uses of that work (i.e., licensing). Indeed, the framers of the constitution believed promotion of the arts was important enough to grant authors and inventors “the exclusive Right to their respective Writings and Discoveries.”[2]

The subject matter protected by copyright generally, extends to literary works, music, dramatic works, pictorial and graphic works, sculptures, songs, and even architecture, just to name a few. The protection lasts for the life of the author plus 70 years for works created after January 1, 1978. What copyright protection does not protect are ideas, methods of operation, principles, and concepts, such as a scientific method or a recipe. There is also no copyright protection for names, titles, and slogans because they contain an insufficient amount of authorship. There lies one of the differences between copyright and trademark: Trademark protects words, phrases, and designs that identify goods and services.

For example, about a decade ago, Dominique Ansel became a viral sensation when his bakery started selling the Cronut, a cross between a doughnut and a croissant. Ansel obtained trademark protection for the word “Cronut” but cannot obtain copyright protection for the process of making the cronut. He would be able to copyright the recipe in the limited sense of protecting the exact pattern of words that make up the recipe, but not the underlying facts or the list of ingredients, generally. (He might consider keeping it confidential and maintaining it as a trade secret, however).

Once upon a time, the author of such a work was required to apply for copyright registration to receive such protection. As of the date of this writing, however, copyright protection exists as soon as the work of authorship is “fixed in any tangible medium of expression, now known or later developed, from which they can be perceived, reproduced, or otherwise communicated, either directly or with the aid of a machine or device.”[3] While it is still a good idea to apply for federal registration, such registration is not a requirement for an author to publish a work with that familiar symbol we’ve all seen before: ©.

But we are in uncharted waters when it comes to how to apply this scope of protection to works generated by AI. The White House issued an “Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence” on October 30, 2023, which provided very little, if any, guidance on the relationship between AI and copyright law.[4] The U.S. Copyright Office (“the Office”) is still trying to make heads or tails of these new developments as well. On August 30, 2023, the Office issued a notice of inquiry stating that it is “undertaking a study of the copyright law and policy issues raised by generative AI and is assessing whether legislative or regulatory steps are warranted.”[5] Regardless of the order of priority, issues concerning the relationship between AI and the scope of intellectual property protection are on the rise. As such, it will largely be up to the courts to set standards in the coming years.

For now, the best we can do is handle one issue at a time. This article will focus on a recent case that may, hopefully, get us a little closer to defining how AI relates to one of the most basic elements of copyright law: authorship. The case to be discussed is Stephen Thaler v. Shira Perlmutter and The United States Copyright Office (1:22-cv-01564) (June 2, 2022). Dr. Thaler attempted to register a copyright for an AI-generated work titled “A Recent Entrance to Paradise,” an image depicting a flowery tunnel entrance that vaguely resembles the style of a classic Monet painting. The Office denied the application on the basis that the work “lacks the human authorship necessary to support a copyright claim” and further stated that Dr. Thaler was not entitled to apply for registration because he listed AI as the author and himself as the owner.[6] In fact, it was not just “AI” that was listed as the author; the author was listed as the “Creativity Machine,” an algorithm Dr. Thaler developed for generating images. Essentially, since Dr. Thaler wrote the code that generated the image, he listed himself as the owner of the copyright because his algorithm generated the image in the capacity of a pseudo-employee under the “work made for hire” doctrine, which essentially says that, when a work is “made for hire” and certain other conditions are met, the author is not the person who created the work, but rather the person who hired the individual who created the work.[7]

Dr. Thaler’s initial application for copyright protection was met with resistance. The Office refused to register the work, stating in support thereof that “this work . . . lacks the human authorship necessary to support a copyright claim.”[8] When a work is denied copyright registration, that decision may be reviewed for reconsideration.[9] Dr. Thaler decided the best course of action in his request for reconsideration was to come out swinging.

Dr. Thaler never attempted to name himself or any other “human” as the author of the work. Instead, he always confirmed that the work was generated by AI and lacked “traditional human authorship” in recognition of the standard that the Office applies when reviewing applications. When the Office denied his first request for reconsideration, Dr. Thaler filed a second request for reconsideration in which he argued that the Office’s requirement that the author of a work be human was both unconstitutional and unfounded in precedent.[10] He further argued that the “work made for hire” doctrine applies to AI-generated works because it recognizes companies as authors.[11]

At this point, I feel that it is important to briefly switch to future tense for dramatic effect and discuss the standard that Dr. Thaler will have to satisfy to prevail. Dr. Thaler’s second request for reconsideration will be examined by a three-member Review Board of the U.S. Copyright Office (“the Board”). For Dr. Thaler to prevail, he must either provide evidence that the work is the product of human authorship or convince the Board to depart from a century of copyright jurisprudence. All parties agree that the work is not the product of human authorship, so there is no need to discuss that requirement. All that is left is to convince the Board to abandon everything they know and believe about human authorship that has been established since the inception of our nation’s copyright laws. No pressure.

Assuming the above subtle foreshadowing was not obvious enough, the Board denied Dr. Thaler’s second request for reconsideration. In response to Dr. Thaler’s constitutional argument, the Board explained that human authorship is an essential element of copyright protection.[12] The Board bolstered their argument with support from the Ninth Circuit and the Compendium,[13] which show consistent findings that works derived from non-human expression are not the kind of works copyright laws were intended to protect.[14]

The Board then discussed the impact of computing technology on copyright.[15] In 1978, the National Commission on New Technological Uses of Copyrighted Works (“CONTU”) concluded that the laws requiring human authorship did not need to be amended and that judicial construction was sufficient to enable protection for works created with the use of a computer. It is essentially the same conclusion that one can obtain copyright on a work generated using a camera. Photographs are entitled to copyright protection because the photographer makes decisions regarding creative elements.[16] However, in order to obtain copyright on a photograph, the camera must still be used by a human: “These court decisions are reflected in the Office’s guidance in the Compendium, which provides examples of works lacking human authorship such as ‘a photograph taken by a monkey’ and ‘an application for a song naming the Holy Spirit as the author.’”[17]

The Board concluded the analysis of the human authorship requirement by affirming that public sentiment has not changed in almost 50 years. In 2020, the United States Patent and Trademark Office (“PTO”) sought public comment on whether a work produced by an AI algorithm without the involvement of a natural person qualifies as a work of authorship.[18] The Board used this public comment report as a final “but don’t just take our word for it” nail in the coffin sealing the rebuttal against Dr. Thaler’s public policy argument: “the vast majority of commenters acknowledged that existing law does not permit a non-human to be an author [and] this should remain the law.”[19] Needless to say, Dr. Thaler’s public policy argument never really stood much of a chance.

As for Dr. Thaler’s assertion of the “work made for hire” doctrine, the Board dismissed his argument with a simple premise: if a machine cannot enter into a contract, as is required for a work-made-for-hire agreement, it cannot author a work under the same.[20] The Board affirmed the decision to not allow copyright protection for Dr. Thaler’s AI-generated work, and his path for relief within the jurisdiction of the Copyright Office ended. However, Dr. Thaler was not done.

Dr. Thaler next appealed the Board’s refusal by way of judicial action under the Administrative Procedure Act (“APA”). With that, we have circled back to Stephen Thaler v. Shira Perlmutter and The United States Copyright Office in which Dr. Thaler argued that the Office’s “denial of copyright registration” was “an arbitrary and capricious agency action and not in accordance with the law.”[21] Dr. Thaler’s arguments were much the same: the Copyright Act allows protection of AI-generated works in the same realm as protection for corporations; the position of the Office is not supported by case law; and the “work made for hire” doctrine permits registration (even though AI cannot execute a binding contract).[22]

Unsurprisingly, Dr. Thaler did not fair any better this next go-around. On August 18, 2023, Judge Beryl Howell issued an order denying Dr. Thaler’s motion for summary judgment and granting the Copyright Office’s cross-motion for summary judgment. The supporting memorandum opinion from Judge Howell noted that Dr. Thaler attempted to complicate the issues by arguing the “viability of legal theories under which a copyright in the computer’s work would transfer to him, as the computer’s owner.”[23] Judge Howell explained that an argument concerning to whom the valid copyright should be registered ignores the fact that no valid copyright ever existed due to the lack of human involvement.[24] This defeated Dr. Thaler’s work-for-hire argument and his other arguments directed to whom this non-existent copyright belongs.

Interestingly, Dr. Thaler did attempt to shoehorn in one issue not previously argued before the Board: that the Court should determine “whether a machine can make something indistinguishable from a person for purposes of copyright protection.” This was too little, too late, because the judicial relief provided by the APA only extends to a review of whether an agency action is found to be “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with the law.”[25] Review is limited to “the grounds that the agency invoked when it took the action” and, as such, there can be no de novo arguments from either party.[26]

Even though the standard of review is limited to what is in the record upon which the agency acted, the Court devotes a significant portion of the discussion to the history and supporting authority to hammer the point home: the author must be human. The laws, as written, were meant to adapt to a changing world and the absence of any human involvement in the creation of the work will not suffice. Sure, a human is responsible for the initial text input that will somewhat guide the program to deliver an output image in line with the user’s request. But at this stage in the game, it’s not enough.

I imagine most people will agree with the decision. One of the fundamental concepts of copyright law is that one cannot copyright an idea. An idea can be expressed in writing or drawings that can then be registered. But to allow registration of an AI output image at this stage would be tantamount to registering an idea while circumventing the requisite human touch. I must admit that my favorite part of the opinion is a playful footnote referencing a Columbia Journal of Law article that states: “The day sentient refugees from some intergalactic war arrive on Earth and are granted asylum in Iceland, copyright law will be the least of our problems.”[27] It emphasizes the mindset of the Court in analyzing Dr. Thaler’s attempt to shoehorn the human authorship requirement of copyright protection and how it relates to what the “big picture” concerning AI is: if AI becomes advanced enough to demand civil rights afforded to humans, we have bigger fish to fry.[28]

Some final food for thought: The opinion went on to emphasize that copyright laws in their current form have been able to cover advancements in technologies throughout time.[29] The Court believes the use of a camera is the use of a mechanical device to reproduce original intellectual conceptions of the author in creating a new type of work. [30] Should an AI program be treated the same as a camera? After all, a camera requires the user to adjust the depth of field, shutter speed, and other variables for it to capture what the user desires. With AI, a narrow, defined text input will yield a more definite output image that may fall in line with the creative vision of a user. Are the images generated by the program too unpredictable, regardless of the specificity of input from the user? Can text prompts and algorithms be enough to bridge the gap from unpredictable to predictable? Needless to say, there have been and will continue to be several legal and societal disruptions caused by the continued advances of AI. Better to face them now because this technology is here to stay.

Endnotes

[1] Christopher T. Zirpoli, Congressional Research Service, Generative Artificial Intelligence and Copyright Law, LSB10922 (September 29, 2023).

[2] U.S. Constitution, Article I section 8, clause 8.

[3] 17 U.S.C.A. § 102.

[4] Joseph R. Biden, Jr., The White House, https://www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/ (October 30, 2023).

[5]Copyright Office Issues Notice of Inquiry on Copyright and Artificial Intelligence, U.S. Copyright Office, NewsNet Issue No. 1017 (August 30, 2023), https://www.copyright.gov/newsnet/2023/1017.html.

[6] Thaler v. Perlmutter, No. CV 22-1564 (BAH), 2023 WL 5333236, at *1 (D.D.C. Aug. 18, 2023) (see Ex. D, Copyright Office Refusal Letter Dated August 12, 2019 at 1, ECF No. 13-4.).

[7] See, e.g., Id.

[8] Perlmutter, 2023 WL 5333236, at *1.

[9] 37 C.F.R. § 202.5(c).

[10] Perlmutter, 2023 WL 5333236, at *1 (see Ex. G, Second Request for Reconsideration at 2, ECF No. 13-7.)

[11] See Id.

[12] See Id. Ex. H, Copyright Review Board Refusal Letter, February 14, 2022 at 1, ECF No. 13-8.

[13] U.S. Copyright Office, Compendium of U.S. Copyright Office Practices § 306 (3d ed. 2017) (provides guidance on institutional practices and principles of law concerning Title 17 of the US Code and Chapter 37 of the Code of Federal Regulations).

[14] See Id. at 4-5, ECF No. 13-8.

[15] See Id. at 5, ECF No. 13-8.

[16] See Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 4 S. Ct. 279, 28 L. Ed. 349 (1884).

[17] See U.S. COPYRIGHT OFFICE, COMPENDIUM OF U.S. COPYRIGHT OFFICE PRACTICES § 313.2 (3d ed. 2021).

[18] See U.S. PATENT AND TRADEMARK OFFICE, PUBLIC VIEWS ON ARTIFICIAL INTELLIGENCE AND INTELLECTUAL PROPERTY POLICY at 19 (2020), available at https://www.uspto.gov/sites/default/files/documents/USPTO_AI-Report_2020-10-07.pdf.

[19] Perlmutter, 2023 WL 5333236, at *1, ECF No. 13-8.

[20] See Id., ECF No. 13-8.

[21] Id. at *2.

[22] See Id. at *3.

[23] See Id.

[24] See Id.

[25] See 5 U.S.C. § 706(2)(A).

[26] Michigan v. EPA, 576 U.S. 743, 758 (2015).

[27] Perlmutter, 2023 WL 5333236, at *4 n.2.

[28] 44 COLUMBIA J. L. & ARTS 383, 408 (2021).

[29] See, e.g. Id. at *3.

[30] See, e.g. Id. at *3.