Contact Us

Disruptive Competition Project

655 15th St., NW

Suite 410

Washington, D.C. 20005

Phone: (202) 783-0070
Fax: (202) 783-0534

Contact Us

Please fill out this form and we will get in touch with you shortly.

Can AI Work Receive a Copyright? Copyright Office Proposes Procedure Update

The U.S. Copyright Office sought public input on revisions to its internal procedures on Friday, proposing language that highlighted the challenges of ascertaining authorship in an era of technology-aided creation.

On Friday, March 15, the U.S. Copyright Office announced the release of an updated public draft to its administrative manual, the Compendium of U.S. Copyright Office Practices, Third Edition, and invited public comment on the 1000+ page document. One critical section, pertaining to machine-aided authorship, was revised to include language from a 1966 Copyright Office report:

The crucial question is “whether the ‘work’ is basically one of human authorship, with the computer [or other device] merely being an assisting instrument, or whether the traditional elements of authorship in the work (literary, artistic, or musical expression or elements of selection, arrangement, etc.) were actually conceived and executed not by man but by a machine.” U.S. Copyright Office, Report to the Librarian of Congress by the Register of Copyrights 5 (1966).

Another noteworthy update was the section on mechanical processes and random selection. In 2017, this text read:

Works made through purely mechanical processes or with an automated selection and arrangement are not eligible for copyright protection. The U.S. Copyright Office will refuse to register a claim in a work that is created through the operation of a machine or process without any human interaction, even if the design is randomly generated. (emphasis added)

The 2019 draft suggests the following changes:

Works made through purely mechanical processes or with an automated selection and arrangement are not eligible for copyright protection without sufficient human authorship. The U.S. Copyright Office will refuse to register a claim in a work that is created through the operation of a machine or process without sufficient human interaction, even if the design is randomly generated. (emphasis added)

The Copyright Office’s proposed revisions to the Compendium follow a January 2019 USPTO seminar, titled Artificial Intelligence: Intellectual Property Policy Considerations (covered by DisCo here). That event also considered questions on the authorship of work by AI.

As the Government grapples with these issues, so have legal academics and experts particularly in evaluating human contribution to AI work.

Is AI-Authored Work Subject to Copyright Protection?

AI-authored work, or work created by AI without assistance from a human, has not been achieved yet, and experts disagree on whether it could ever be. James Grimmelmann, Professor of Law at the University of Maryland, says in his article, There’s No Such Thing as a Computer-Authored Work And It’s a Good Thing, Too:

I would like to talk about computer-authored works I would like to, except that they don’t exist.

However, if it were possible for AI to create work completely unassisted by humans, law in the U.S. suggests that the work would not be subject to copyright protection. The infamous “monkey selfie” dispute paved the way for clarification of non-human authorship in copyright. In 2011, an Indonesian monkey took a selfie after a wildlife photographer left his camera unattended. PETA brought the case on behalf of the monkey, arguing that a monkey could hold a copyright. However, the federal district court dismissed the case (which was later affirmed by the Ninth Circuit), ruling that a monkey lacked standing under the Copyright Act. In 2014, the U.S. Copyright Office updated its Compendium, stating that non-humans cannot hold a copyright, and specified “photography by a monkey” as an example of work that lacks human authorship.

Because non-humans cannot hold a copyright, the Copyright Office will also refuse works “produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author.” Thus, AI-authored work would most likely fall under this category; without human intervention, the AI itself would not be able to hold a copyright.

Yet despite the monkey selfie litigation, some commentators distinguish animals from the rapidly evolving capabilities of AI and envision the potential need to consider copyright protection for AI instruments as they become smarter, more creative, and less dependent on human assistance. Additionally, the current status quo of only granting humans copyright protection may become more nuanced due to the blurring of lines of what it means to be a legal person. For example, ‘Sophia the Robot’ obtained citizenship status in Saudi Arabia (and therefore “legal personhood”), and the EU is considering the possibility of robots as “electronic persons”.

Is AI-Generated Work Subject to Copyright Protection?

Last October, an AI-generated art piece titled Portrait of Edmond Belamy sold for $432,000 at a Christie’s auction. Despite all the hype about the Belamy art piece as AI “creating” work, Belamy is better categorized as AI-generated. AI-generated work is work in which AI augments or assists humans. Hugo Caselles-Dupré, the technical lead at Obvious and one of the makers of Belamy, described the present state of AI as such:

Right now it’s like a computer program that can do something that we find impressive, but it is not really that impressive in the end, because if you think it’s some kind of robot with a human-like mind, okay, it’s really, really impressive, but it’s not.

Jason Bailey, a digital art blogger behind Artnome, echoed that sentiment, and argued that “anyone who has worked with AI and art realizes algorithms are tools, not active collaborators or autonomous agents.”

Who could hold a copyright for AI-generated work? Professor Annemarie Bridy, a Professor of Law at the University of Idaho, believes that the work-for-hire doctrine is the appropriate framework to evaluate this question. The work-for-hire doctrine, codified under 17 U.S.C. § 101, provides that “the employer or other person for whom the work was prepared is considered the author”. However, since there is no codified correlation in copyright law between the programmer of an AI and the AI itself, AI-generated work would not fall under the traditional definition of work-for-hire. Therefore, Professor Bridy suggests that the statutory definition could be amended to incorporate the definition of computer-generated work from other jurisdictions. For example, in the United Kingdom, section 9(3) of the Copyright, Designs and Patents Act states:

In the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken.

If the U.S. were to adopt and incorporate this definition into its work-for-hire doctrine, then in cases of AI-generated work, the author would similarly be the person by “whom the arrangements necessary for the creation of the work are undertaken” and could thus argue for copyright protection.

If AI could produce work without human intervention at all, then perhaps no one could hold a copyright to AI-authored work. On the other hand, AI-generated work is already prevalent in the art world. The proposed changes to the 2019 draft Compendium may be seeking to clarify perceived gaps in current law by allowing for copyright protection in AI-generated work as long as there is “sufficient human authorship.”

Intellectual Property

The Internet enables the free exchange of ideas and content that, in turn, promote creativity, commerce, and innovation. However, a balanced approach to copyright, trademarks, and patents is critical to this creative and entrepreneurial spirit the Internet has fostered. Consequently, it is our belief that the intellectual property system should encourage innovation, while not impeding new business models and open-source developments.