Generative AI and Recent Copyright Developments

The recent expansion of the scope and capabilities of generative artificial intelligence (AI) tools and platforms has introduced a number of legal challenges. These challenges can represent a double-edged sword, both for those seeking copyright protection and those accused of copyright infringement involving works and materials created using generative AI tools and platforms. As the applications of generative AI tools and platforms continue to expand and accelerate, the United States Copyright Office (USCO) and pending litigation in U.S. district courts have already started to weigh in on critical issues concerning copyright ownership, registration and infringement.

Generative AI and Copyright Protection

A recent decision by the USCO shows the challenges that may be faced by those seeking copyright protection for works that include components or features that were created by a generative AI tool. In this case, the USCO initially cancelled the copyright registration of a comic book “Zarya of the Dawn,” which included images created by a generative AI tool.¹ The USCO subsequently found, however, the portions of the comic book attributable to the efforts of the human artist, Ms. Kristina Kashtanova, were still entitled to copyright protection. More specifically, the text, which Ms. Kashtanova authored, and the arrangement of the AI-generated images, which she also performed, could be registered for copyright protection. By drawing distinctions between the text, the images and the arrangement of the images, the USCO drew boundaries illustrating the portions of works that can—and those that cannot—be afforded copyright protection when an author is assisted by generative AI. As a result, the USCO re-registered the comic book, excluding from the registration, the images generated by AI.

In reaching its decision, the USCO relied on the United States Copyright Act, which provides that “original works of authorship fixed in any tangible medium of expression” are entitled to copyright protection.² The phrase “works of authorship” was deemed to require creation by a human author, while the term “original” requires independent creation and sufficient creativity.³

In accordance with these statutory standards, the USCO determined the selection and arrangement of the AI-generated images is protectable as a “compilation.” In particular, these exclusive efforts by Ms. Kashtanova satisfied the human authorship requirement, and the arrangement of the images for the purpose of telling a story satisfy the statutory requisite for creativity.

In contrast, the USCO determined that the AI-generated images themselves did not qualify for copyright protection because they are not individual works of human authorship. The USCO examined the process through which generative AI creates these images, emphasizing that process used by the generative AI platform is not the same as a human author. Specifically, an individual enters a text prompt into the generative AI platform, which outputs images the individual may use as intermediates that can be further refined through additional text prompts to reach a final image. Based on this iterative process, Ms. Kashtanova argued that she is the author of the images because she “guided” their structure and content through trial and error.⁴ Ultimately, however, the USCO rejected this argument, finding that Ms. Kashtanova lacked sufficient control over generated images to be treated as the inventive or master mind behind them.⁵

Notably, the approach taken by the USCO is consistent with that of the United States Court of Appeals for the Federal Circuit in the patent context for AI-based inventions. Much like USCO’s interpretation of the Copyright Act, the Federal Circuit held the United States Patent Act requires the named inventors of a patent to be a human being.⁶

Generative AI and Potential Copyright Infringement

Although AI’s assistance does not prevent an individual from obtaining a limited registration from the USCO, there may be other legal challenges to consider. As one example, several generative AI platforms are at the forefront of pending litigation relating to copyright infringement.

In January 2023, several companies offering generative AI tools were named as defendants in a proposed class action lawsuit alleging copyright infringement. The proposed class consists of artists alleging infringement on the basis that the platforms are trained using the artists’ copyrighted works without first receiving a license or other consent from the artists. Recently, the presiding judge indicated at oral argument that the complaint was likely to be dismissed, though the plaintiffs would be granted leave to file a new complaint to add additional facts in support of the plausibility of the claims.⁷

Given the legal industry is at the dawn of potential legal claims and challenges involving generative AI, we expect the generation of legal decisions that will soon affect the application of copyright law (and other intellectual property protections) to this still-developing technology.

_{1 Letter from U.S. Copyright Office to Van Lindberg, counsel to Kristina Kashtanova, regarding Zarya of the Dawn (Feb. 21, 2023), https://copyright.gov/docs/zarya-of-the-dawn.pdf.}

_{2 17 U.S.C. § 102(a).}

_{3 Supra note 1, at 3-4.}

_{4 Letter from Van Lindberg to the U.S. Copyright Office (Nov. 21, 2022), https://copyright.gov/docs/zarya-of-the-dawn.pdf.}

_{5 Supra note 1, at 9.}

_{6 David C. Vondle & Megan R. Mahoney, Federal Circuit Confirms “Inventor” Must be Human, Not AI, Akin (Aug. 11, 2022), https://www.akingump.com/en/insights/alerts/federal-circuit-confirms-inventor-must-be-human-not-ai.}

_{7 Bonnie Eslinger, Back to Drawing Board for Artists’ IP Suit Against Stability, LAW360 (July 19, 2023), https://www.law360.com/articles/1701543/back-to-drawing-board-for-artists-ip-suit-against-stability-ai.}