“Should the USCO update the application process to account for technological advancements for machine-generated works? If it does, the question remains as to where on a so-called ‘creation-generation spectrum’ the line is drawn for whether or not a work is eligible for registration.”
On June 2, Dr. Stephen Thaler filed a complaint in the U.S. District Court in Washington, D.C. naming as defendants both the United States Copyright Office (USCO) and Shira Perlmutter, in her official capacity as Register of Copyrights and Director of the USCO. The complaint marks the start of a new phase of Thaler’s attempts at obtaining a copyright registration for “A Recent Entrance to Paradise,” an AI-generated work that is the output of Thaler’s AI system known as Creativity Machine. Thaler is requesting the district court issue an order that would require the USCO to set aside the Review Board’s decision and reinstate the application for registration of the work. Thaler is also seeking an award of reasonable attorneys’ fees and costs. The case is Stephen Thaler v. Shira Perlmutter and The United States Copyright Office (1:22-cv-01564) (June 2, 2022). Thaler is represented by Geoffrey “Geoff” Neri and Ryan Abbott (pro hac vice), of Brown, Neri, Smith & Khan LLP in Los Angeles, CA. No representatives have yet been identified for the USCO.
In February 2022, the Review Board of the USCO issued an opinion letter in which it affirmed the USCO’s previous decisions to deny registration of the AI-generated work. “If the U.S. Copyright Office upholds the refusal to register following a request for reconsideration,” according to Section 1706 of the Compendium of the U.S. Copyright Office Practices, Third Edition, “an applicant may appeal that decision under the Administrative Procedure Act (“APA”) by instituting a judicial action against the Register of Copyrights.”
“This is a pro bono legal test case being done as part of The Artificial Inventor Project to help provide guidance to those involved in generating AI-Generated Works and promote a discussion about what rules can be help us maximize the benefits of frontier technologies like AI,” shared Dr. Ryan Abbott, a Partner at Brown Neri Smith Khan, LLC, the firm representing Thaler, in a written response to a request for comment. “The public should take away [from this] that continued advances in AI are going to be highly disruptive socially, economically and legally.”
The complaint sets forth several arguments supporting what Thaler identifies as “arbitrary and capricious agency action” that doesn’t align with the law. The arguments largely mirror those previously made by Thaler during the administrative appeals process within the USCO, with some additions that serve as indirect responses to the Review Board’s letter.
Thaler starts by asserting that the plain language of the Copyright Act allows for the protection of AI-generated works under a concept similar to protections afforded to corporations and other non-human entities (to meet the “authorship” requirement). An argument is also made regarding an overall low bar to meet the originality requirement.
Thaler also asserts there is a lack of case law supporting the USCO’s position during its reviews and denials that AI-generated works are ineligible for copyright protection. Thaler points to the lack of any knowledge with regards to advanced technologies, such as AI, during the 1800s when the Supreme Court decided Burrow-Giles Lithographic Co. v. Sarony, a case referenced by the Review Board in its letter. Instead, Thaler argues “our courts have a long history of purposive interpretation of the Act in light of technological evolution.”
Thaler proposes that the Turing Test, developed by Alan Turing in 1950, should be applied by courts. More specifically to copyright law, Thaler proposes that “[t]he real question is whether a machine can make something indistinguishable from a person for purposes of copyright protection.” To which the answer is argued by Thaler to be “yes” with respect to the “A Recent Entrance to Paradise.” Thaler also spends time in the complaint distinguishing two Ninth Circuit cases, Naruto v. Slater (the monkey selfie case) and Urantia Foundation v. Maaherra (a case involving a book partially created by spiritual beings).
Another argument put forth by Thaler is that the work may be classified under the work made for hire doctrine, an argument that was previously made to the USCO. In the complaint, Thaler acknowledges the Review Board’s reply argument that an AI is not an employee or an independent contractor that can execute a contract. However, Thaler argues AI “functionally behaves” as such, and therefore should be afforded similar status.
The argument regarding the Copyright Act’s originality requirement is interesting, and one where Thaler should expand on the technical aspects of the Creativity Machine. Thaler includes what can be assumed to be a supporting quote from Feist Publications, Inc. v. Rural Telephone Service Company, Inc., “Original, as the term is used in copyright, means only that the work was independently created by the author (as opposed to copied from other works), and that it possesses at least some minimal degree of creativity.”
However, if AI models are only as good as the data used to train them (and such data can be assumed to be pre-existing works), Thaler’s argument is missing the connection that explains how the output of a model doesn’t automatically fall within the bucket of being “copied from other works” simply based on the nature of how AI technologies work. Is there room to apply the concept of human learning, or what Thaler refers to as “human-centric mental activity”, to AI model training?
“The fact USCO has no way of knowing whether a work is AI-generated,” notes Abbott, “is one more reason why an antiquated human authorship requirement in not an appropriate policy.”
Additionally, Thaler highlights a growing problem for the USCO and the current process for registration of works, namely that the entire system relies on an honor system for disclosure of a work having been generated by AI software or similar algorithm. “[I]f Thaler had submitted the same AI-Generated Work with his company listed as the author,” argues Thaler in the complaint, “USCO would have granted his company a registration, and no one would have known the work was AI-Generated.”
The Review Board, in a footnote to the Letter, acknowledged criminal penalties for “[a]ny person who “knowingly makes a false representation of a material fact in the application for copyright registration . . . or in any written statement tied in connection with the application.” 17 U.S.C. § 506(e). The Review Board noted further that “[a]pplicants who mislead the Office do so at their peril.“ For the uninformed applicant, perhaps a more likely outcome would be the cancellation of a registration where the “material deposited does not constitute copyrightable subject matter” or “the claim is invalid for any other reason.” 17 U.S.C. § 410(b); see also Compendium (Third) § 1807.
“We argue the law is flexible enough to currently accommodate AI-generated works,” Abbott says, “but if courts ultimately disagree with us, then it would be to legislatures to adjust the system to ensure it results in good outcomes.”
(Spectrum Graphic Credit: Franklin Graves)
The Way Forward
Should the USCO update the application process to account for technological advancements for machine-generated works? If it does, the question remains as to where on a so-called “creation-generation spectrum” the line is drawn for whether or not a work is eligible for registration. The creation-generation spectrum, illustrated in the above graphic, recognizes a ‘tool’ as being any non-human mechanism, machine, or medium, whether technical (such as an algorithm) or non-technical (such as a paint brush), that aids a human author in the creation and/or generation of a work. The far-left side of the spectrum identifies works created without the use of any tools, such as an ephemeral vocal or dance performance. The far-right side of the spectrum identifies the theoretical concept of the singularity – the point at which AI is self-aware.
It is at some point along this spectrum where jurisdictions around the world will need to identify and develop policy and practical guidance for creators (and inventors) when it comes to the creation-generation gap in copyright law.
Not having clear guidance on where on the creation-generation spectrum the line is drawn for when a work becomes unprotected by copyright laws puts multi-billion dollar markets, such as NFTs, at risk of being worthless or subject to other known, or as yet unestablished, proprietary rights systems.
Meanwhile, Thaler’s Patent Fight Plays Out at CAFC
Separately, the U.S. Court of Appeals for the Federal Circuit today heard oral arguments in Thaler v. Vidal, where the judges are considering whether Thaler’s AI machine, “Device for the Autonomous Bootstrapping of Unified Sentience,” or DABUS, can be considered an inventor under the U.S. Patent Act.
Abbott, who is also Thaler’s counsel in this case, argued that the term “individual” needs a broader meaning as it’s applied to an inventor and that the Patent Act is agnostic as to how a patent is generated. Abbott warned against closing off the possibility of invention by machines or companies because it will inhibit innovation, as companies won’t take advantage of AI if they can’t get patent protection. But the USPTO countered that the Patent Act defines inventors as individuals and then uses gendered personal pronouns (“himself” and “herself”) to refer to those individuals, and that the problems raised by Thaler must be tackled by Congress.