Federal Court Upholds Copyright Office Refusal to Register AI Output
On March 16, 2023, the U.S. Copyright Office published a policy statement saying that the Copyright Office would generally not register copyrights in the output of artificial intelligence. The Copyright Office invoked the traditional requirement for a work to be copyrightable that the creative authorship must be that of a human author.
Stephen Thaler owns a computer system he calls the “Creativity Machine,” which he claims generated a piece of visual art of its own accord. Thaler sought to register the work for copyright, listing the computer system as the author and explaining that the copyright should be his as the owner of the machine. The Copyright Office, in accordance with its published policy, denied the application on the grounds that the work lacked human authorship. Thaler challenged that denial, culminating in a lawsuit filed in the federal court for the District of Columbia against the Copyright Office, titled Thaler v. Perlmutter. On August 18, the court upheld the Copyright Office’s position.
There is no doubt that Thaler’s picture would have been registered if it had been created exclusively by a human being. The Copyright Office refused to register copyright because the picture was created by artificial intelligence, even though it was Thaler who built the AI. The court agreed with the Copyright Office that, to be entitled to copyright, creative authorship in a work had to be done by a human author rather than a mechanical process, saying:
“Author,” in its relevant sense, means “one that is the source of some form of intellectual or creative work, [t]he creator of an artistic work; a painter, photographer, filmmaker, etc.” Author, MERRIAM-WEBSTER UNABRIDGED DICTIONARY, (last visited Aug. 18, 2023); Author, OXFORD ENGLISH DICTIONARY, (last visited Aug. 10, 2023). By its plain text, the 1976 Act thus requires a copyrightable work to have an originator with the capacity for intellectual, creative, or artistic labor. Must that originator be a human being to claim copyright protection? The answer is yes.
It should not be assumed, however, that this decision closes the door to copyright in works assisted by or even largely created by artificial intelligence. Possibly the most interesting part of the opinion is in footnote 1:
In pursuing these arguments, plaintiff elaborates on his development, use, ownership, and prompting of the AI generating software in the so-called “Creativity Machine,” implying a level of human involvement in this case entirely absent in the administrative record. As detailed, supra, in Part I, plaintiff consistently represented to the Register that the AI system generated the work “autonomously” and that he played no role in its creation, see Application at 2, and judicial review of the Register’s final decision must be based on those same facts. (emphasis added)
This footnote leaves open the possibility that a designer or creator of AI could claim copyright in its output by proving a sufficient degree of control or influence over the AI’s output. Much can be done with AI to direct or influence its output. It may be that Thaler did not do enough to control or determine the AI output to be able to make a serious case that the creative authorship was his, merely assisted by AI, and did not pursue that argument for that reason.
There is practically no copyrightable work that is made without a machine or other tool. Photographs, for example, are copyrightable even though they are the output of a machine. Modern cameras may be equipped with automatic features, such as autofocus or automatic selection of shutter speed and exposure. The human photographer, however, may manipulate the automatic settings to achieve specifically desired effects and the photographer has complete control of the picture’s subject matter. The photographer’s creativity in setting or selecting the subject and scene and in setting the camera to produce artistic effects makes the picture copyrightable.
Similarly, word processing software contains spelling and grammar correction tools, thesaurus features to aid word selection, and more recently, translation modules, but the overall selection of words remains exclusively with the author, who typically accepts or rejects the computer’s suggestions.
These examples are relatively simple, because computer software is being employed to capture and transcribe the author’s original expression. Other software programs, however, have long performed much more complicated tasks than merely “transcribing” a human user’s inputs or offering editing suggestions. Some software applications will generate complicated output using very few human-controlled inputs. Music generation software, for example, can enhance a simple melody with computer generated chords, rhythms, or accompaniments matching particular musical styles.
Legal scholars have noted the sliding scale of control that a human user of an automated software system may exert over the software’s output. See e.g., Jane Ginsberg and Luke Ali Budiardjo, “Authors and Machines,” 34 Berkeley Tech. L.J. 3433 (2019). This article distinguishes between tools whose outputs reflect the creative contributions of their users; fully generative machines, whose outputs reflect the creative contributions of their designer; and tools which combine the two.
Consistent with the framework described by these authors, AI systems, unlike cameras and word processors before them, can be configured to be “completely autonomous.” Indeed, Thaler stated in his copyright application that his “Creativity Machine” autonomously generated the image in issue. An AI system can also be configured to act based on combinations of creative controls, namely the “development, use, ownership, and prompting” of the AI generating software that Thaler subsequently argued in support of the copyrightability of that same image. As Professor Jane Ginsberg suggests, those combinations of creative controls may be enough to satisfy the Copyright Act’s standards of authorship, even though the machine is the instrumentality through which the “developer, user, owner and prompter” of the AI system creates its output.
The court’s decision in Thaler v. Perlmutter and the Copyright Office’s position on copyrightability of the outputs of AI are recent developments. But the framework for deciding when a combination of creative controls of an AI system are enough to result in a copyrightable output is not new. In March 2020, Moses Singer hosted Professor Ginsberg and other AI experts for an “A.I. Micro-summit” at which Professor Ginsberg presented her then-recently published paper. As Professor Ginsberg explained then, and her article cited above notes, “Asking whether a computer can be an author therefore is the ‘wrong’ question; the ‘right’ question addresses how to evaluate the authorial claims of the humans involved in either preparing or using the machines that ‘create.’”
Professor Ginsberg’s analytical framework would allow courts to grant a human copyright in the output of an AI system if that the human exerted sufficient control over that output. Examples: human authors can customize or rearrange AI-generated blocks of text; human authors can instruct an AI system how to refine or revise AI’s initial output. “Setting” AI could be far more creative than setting the various settings on a camera. It is not hard to imagine that a combination of choosing the training data, instructing the AI how to interpret the training data, and giving detailed instructions (prompts) to the AI could so strongly influence the output to entitle the AI’s creator (or user) to copyright protection.
The case of works resulting from human involvement in the details of AI output remains to be made.
As noted above, the court’s decision in Thaler v. Perlmutter leaves open the question of how much human involvement is needed to make AI output eligible for copyright protection. That decision is surely not the last word on copyrightability of works generated by AI. Thaler himself has already announced that he will be appealing the district court’s decision. It also will be worth watching whether Congress will legislate a standard for AI output copyrightability.
Finally, in an August 30, 2023 Notice in the Federal Register (88 FR 167, p. 59942), the Copyright Office announced that it is undertaking a study of the copyright law and policy issues raised by AI systems to determine whether legislative or regulatory steps are warranted in this area, and is seeking public comment on these issues.
David Rabinowitz, Toby Butterfield & Milton Springut are partners at Moses & Singer LLP, a practice who believe in investing heavily in understanding their clients’ businesses and developing close working relationships with them. David focuses in the substantive areas of financial industry litigation, including corporate trusts and letters of credit, trusts and estates, intellectual property, contracts and employment. Milton focuses on intellectual property litigation and counseling. He litigates and prosecutes patents in the scientific disciplines, including electrical and electronic systems, computer hardware and software, and business systems.
This article is intended for informational purposes only, and doesn't constitute tax, accounting, or legal advice. Everyone's situation is different! For advice in light of your unique circumstances, consult a tax advisor, accountant, or lawyer.