Me, Myself, and (A)I: Copyright Office to Focus on AI Authorship

This content has been archived. It may no longer be accurate or relevant.

From Lexology:

According to a recent interview in December 2022, the U.S. Copyright Office (the “Office”) signaled that it would focus in 2023 on “legal grey areas” surrounding copyrightability of works generated in conjunction with artificial intelligence (“AI”) tools. While the agency is standing by its conclusion that copyright cannot be registered for a work created exclusively by an AI, according to Shira Perlmutter, the Register of Copyrights and director of the Office, the Office is considering the issue of copyright registration for works co-created by humans and AI. The focus is timely given recent monumental leaps in natural language processing and image, text, and code generative AI.

The ‘Human Authorship Requirement’

In 2018, a computer scientist named Stephen Thaler submitted an application to the Office to register a copyright in artwork entitled A Recent Entrance to Paradise (the “Work”). The application identified the “Creativity Machine” as the sole author of the Work, with Thaler listed as a co-claimant. In the application, Thaler noted for the Office that: (1) the Work was automatically created by a computer algorithm running on a machine; and (2) he was seeking to register the computer-generated work as a work-for-hire to Thaler as the owner of the Creativity Machine.

The Work was generated in 2012 by “Device for the Autonomous Bootstrapping of Unified Science,” or DABUS, an AI system developed by Thaler. In an August 2019 letter, the Office refused to register the copyright, finding that DABUS lacked the “human authorship necessary to support a copyright claim.” Thaler submitted requests for reconsideration in September 2019, May 2020, and February 2022, suggesting that “the Office’s human authorship requirement is unconstitutional and unsupported by case law,” which the Office rejected for lack of “traditional human authorship.” In the 2020 determination, the Office again concluded that the Work “lacked the required human authorship necessary to sustain a claim in copyright” because Thaler had “provided no evidence on sufficient creative input or intervention by a human author in the Work.” The Office also noted that it would not “abandon its longstanding interpretation of the Copyright Act, Supreme Court, and lower court judicial precedent that a work meets the legal and formal requirements of copyright protection only if it is created by a human author.”

In stating its conclusion in its 2022 denial of reconsideration and affirming the initial refusal to register the copyright claim in the Work, the Office affirmed the prevailing legal rule for a century that copyright law protects “the fruits of intellectual labor” that “are founded in the creative powers of the [human] mind.” U.S. Copyright Office, Compendium (Third) of U.S. Copyright Office Practices § 306 (quoting Trade-Mark Cases (1879)). Additionally, the Office noted (as a corollary) that it does not register works produced by a machine or mere mechanical process that operates without any creative input or intervention from a human author. Accordingly, Thaler had to either: (1) provide evidence that the Work was a product of human authorship; or (2) provide an argument convincing the Office to depart from the prevailing legal rule, and the Office noted that Thaler accomplished neither. Given the concerns that IP policy is lagging behind technology, the Office “is exploring open questions on copyright registration for works created by humans in conjunction with AI.”

Beyond the Thaler Determination: The Next Frontier of Co-Creation

In 2022, OpenAI released the AI text generator ChatGPT and the AI image generator DALL-E 2. ChatGPT is a natural language processing tool that allows users to have human-like conversations, receive assistance with tasks such as composing code, essays, and emails, and automate otherwise mundane operations. DALL-E 2 is a language model that creates images from user-submitted text captions for a wide range of concepts expressible in natural language. These tools and other generative AI tools made available by OpenAI’s competitors have raised questions about the protectability of works created by humans in conjunction with AI.

In the December 2022 interview, Director Perlmutter, noting that the Office applied existing case law holding that human authorship is a prerequisite for copyright protection in the Thaler determinations, states that the more challenging issues will involve a component of human creativity. Once the human element is involved, the question becomes whether it rises to the level of authorship under existing case law. A complication is that the Office must rely solely on the facts presented in an application for copyright registration as it currently has no way to verify whether an AI system has been appropriately credited in a copyright application.

Recently, the Office began proceedings to cancel a copyright registration it initially issued to Kristina Kashtanova for her AI-assisted graphic novel “Zarya of the Dawn.” The novel includes AI-generated images which Kashtanova created using Midjourney, an AI software platform that generates image outputs based on text inputs. The Office commenced its move to cancel the registration after public disclosures that the novel was in part composed of AI-generated images that were not disclosed in the copyright application. The Office can move to cancel a registration if the authorship is insufficiently creative or the work does not contain authorship subject to copyright on the grounds that the registration is invalid under the applicable law and regulations, 37 C.F.R. § 201.7.

In addition to the question of copyrightability, other emerging AI-related questions that the Office and the Courts are likely to increasingly face include: (1) what level of creative input must exist for an AI-generated work to constitute human “authorship” to support a copyright claim; and (2) should AI-created works that are derived from analyzing existing works (such as those used for training data) or styles require a license from the original author to avoid a claim of copyright infringement, and to what extent do principles of fair use apply?

Link to the rest at Lexology

PG notes that the authors of the OP are associated with the Intellectual Property Group at Davis Wright Tremaine LLP, a large law firm with offices in Seattle and a great many other locations.

11 thoughts on “Me, Myself, and (A)I: Copyright Office to Focus on AI Authorship”

  1. The use case I see for non-trivial uses of AI text generators is first drafts. I can imagine putting in a fact set as the prompt, letting the AI spit out some text, then reviewing it for accuracy while editing it. I am unpersuaded that this will result in good writing, but good enough writing is plausible, and often all that anyone is looking for. The question is how much human involvement is enough to rate copyright? If I do the work of researching the prompt and cleaning up the result, am I a writer or an editor?

      • I don’t think it does, F.

        I also am not aware of anyone who thinks the copyright office will object if you use an AI for various parts of a work to create alternative expressions of ideas. You might take the AI pieces as-is or modify them to suit your fancy.

        I’ll throw another cat amidst the pigeons. I am not aware of any obligation an author has to disclose the tools she/he/they used to create an original work. Word, WordPerfect, Google Docs, ChatGPT are each a tool that can be used by an author to create an original written work.

        • That was my assumption as it should not matter how the sausage is made; just that the proper originator gets the copyright. I use “originator” because work for hire is a thing and so is “AI”.

      • It has been Copyright Office policy to ignore all aspects and implications of any distinction between “process” and “product” since 1910. Yes, there is a memo. (No, I didn’t write it, I’m not quite that old.)

    • Agreed on your first draft opinion, R.

      However, I also think AI in its present state might be useful for alternatives to a paragraph that doesn’t seem quite right for some reason the author can’t figure out.

  2. I would note that the USCO backtracked (in November of last year) on deregistering the “Zarya of the Dawn” copyright – and rightly so.

    A comic book is a collection of images, arranged in a sequence, and (usually) with accompanying text to tell a story. “Zarya of the Dawn” certainly is a comic book of this exact type. It requires a great deal of human creativity.

    About the only way that I could see rejection for “insufficient human creativity” would be if the story was generated by a text AI, the pieces of which were submitted to the graphic AI. Not the case there, at all.

    • Even then some human had to be puppeteering the software. As long as the same person gets the copyright there is no need for pearl clutching.

      • What if they were artificial pearls? <vbeg>

        All seriousness aside, that conflates the distinct “is this copyrightable” and “is the claimant entitled to claim the copyright” issues. (Compare the work made for hire doctrine…)

        • They’re not griping over the content (“Is it intrinsically not copyrightable.”) but over the process (“It has cooties becsuse they didn’t hire an artist to make the illustrations!”)

          They started it.

          (Marvel hires writers, artists, and editors to put out their comics. The creatives get a check and a credit. Marvel gets the copyright. Joe Indie uses software to put out his comic and the copyright is in question? Methinks I hear Ned Ludd off in a corner whispering to the Copyright office. Or is it Mickey…?)

Comments are closed.