Login | November 18, 2025
Protecting artificial intelligence with some old-fashioned know-how
DORIS ESTELLE LONG
Law Bulletin columnist
Published: June 1, 2018
Open the pod bay doors, HAL.”
Whether your view of artificial intelligence is shaped by “2001: A Space Odyssey,” “The Terminator” or “Westworld,” there is little doubt that the imagination of earlier eras is rapidly becoming reality.
From monkey selfie “authorship” to new definitions of “work for hire” that recognize robotic “employees,” scholars are debating radical changes to copyright law to accommodate AI. Although copyright should be amended to reflect the special challenges of AI-generated works, radical reforms reflecting our growing “romance with the machine” are both unnecessary and unwise.
Instead, a process-based solution, created by Great Britain to deal with a similarly shattering technological development from the last century — computer software — already exists.
Adapting this solution is relatively simple. It also creates an equitable balance between protection and access that ensures the benefits of the AI revolution are fairly available to all.
AI-enabled machines and processes, with their critical functions of “deep learning” and “predictive analytics,” are revolutionizing the economy, including the creative arts. From oil paintings such as “The Next Rembrandt,” to songs, news articles and a short novel that made it past the first round in a Japanese literary competition, AI has entered a new area of content creativity.
According to published reports, “The Next Rembrandt” was created by an AI-enabled computer into which were input the 346 known works by Rembrandt van Rijn, facial-recognition software and a command to create a new work. Theoretically, there are three potential authors of this AI-generated, down-the-line, work:
The author of the AI software because it is a derivative work.
The programmer who selected and input the data because it is a compiled work.
The nonhuman computer.
Today, it seems absurd to suggest that anyone other than the human author of a novel written using a third-party word processing software would own its copyright. Authorship of a work created where the computer is used simply as a production tool, so-called computer-aided creation, is well-established. Yet in the early years of computer processing, the issue was hotly debated. It has re-emerged in today’s AI debates.
In the United States, specific reforms to cover the challenges arising from computer software were limited to a definition for “computer program” and a section establishing special software adaptation, archive and repair rights. (17 U.S.C. Sections 101, 117).
By contrast, in the 1988 Copyright, Designs and Patent Act, Great Britain, using a process-based approach, created three tiers of authorship for “computer-generated” works.
These tiers reflect the previously identified three potential authors: (1) the person who “creates” the work; (2) the person “by whom the arrangements necessary for the creation of the work are undertaken”; and (3) the machine if the work is “generated by [a] computer in circumstances such that there is no human author of the work.” (CDPA, Section 9(1), 9(3) & 178).
In the only reported case under the act dealing with this issue, Nova Productions Ltd. v. Mazooma Games Ltd. (2006), the British court held that individual frames shown on a screen when playing a computer game were computer-generated works.
It further held that the individual who made “the arrangements necessary for the creation of the work” was the programmer who “devised the appearance of the various elements of the game and the rules and logic by which each frame is generated and … wrote the relevant computer program.”
The court held that the player of the game, by contrast, was not an author because “his input is not artistic in nature and he has contributed no skill or [labor] of an artistic kind.”
Extrapolating on the court’s analysis, “the necessary arrangements” required to secure authorship to the “arranger” are analogous to the obligations applied in connection with databases and other compiled works.
In Great Britain, originality in such works is based on the presence of sufficient independent skill, labor or judgment to result in a qualitative alteration of the underlying works. (The Newspaper Licensing Agency Limited v. Meltwater Holding BV). In the United States, authorship requires the exercise of judgment and choice in the selection or arrangement of such works, or both (Kregos v. Associated Press).
Based on this approach, authorship of “The Next Rembrandt” becomes relatively predictable. The individual who created the AI program used to produce “The Next Rembrandt” is the “author” of that program. If “The Next Rembrandt” is within the range of foreseeable choices established by the AI program, then she or he is also the author of that down-the-line work.
If, however, the creator of the program is not the author of “The Next Rembrandt,” then there is no protected human author. Even under the relatively low standards for originality in compiled works, it is doubtful that inputting all publicly acknowledged paintings by Rembrandt to create a compiled new work reflecting his artistic techniques would rise to the necessary level of creative selection for authorship (William Hill (Football) Limited v. Ladbroke (Football) Limited; Feist).
Consequently, “The Next Rembrandt” would qualify as an “AI-generated work” for which no human author and, therefore, no current copyright protection exists.
Although international treaties and most domestic laws currently anticipate authorship in the form of a human agent, they do not prohibit the grant of copyright for nonhuman authored works. But at a time when the practical scope of copyright is limited by digital piracy and expanding exceptions in the name of information and cultural access, there seems no defensible policy for expanding copyright to include an exponentially greater array of protected works by nonhumans.
To the contrary, solutions such as extending the “work for hire” doctrine to cover nonhuman “employees” make the ownership of creative devices the critical battleground.
This places debates over device “creativity” and “bring your own device” policies in the crosshairs of copyright policy. Relying on the fair-use doctrine to resolve access problems created by an expanded copyright regime only places greater strain on an already over-taxed, unpredictable exception.
When current alternatives, such as patents and trade secrets, exist to encourage machine-driven innovation, there appears little reason to place copyright law into such turmoil.
AI-generated works open new vistas for creativity. The answer to the challenges they pose to copyright fortunately lies in the past — if we are wise enough to recognize it.
Doris Estelle Long is the president of Doris Long Consulting, specializing in U.S. and international IPR and information security issues; a screenwriter and producer for VeraKen Productions; and a law professor emeritus at The John Marshall Law School. She has served as a consultant on IPR issues for diverse U.S. and foreign government agencies, including as attorney adviser in the Office of Legislative and International Affairs of the USPTO. She can be reached at prof.doris.long@gmail.com.