AI and Copyright Law: Recent Developments in AI-Generating Infringement Suits

The rapid advancement of artificial intelligence (AI) technologies has ushered in a new era of innovation across various sectors. However, this progress has also led to significant legal challenges, particularly in copyright law. These cases underscore the growing tension between AI development and intellectual property (IP) rights, highlighting the need for clear legal frameworks to address these emerging issues.

On June 24, 2024, Sony, Universal, Warner Music, and several other record labels, part of the Record Industry Association of America (RIAA) (collectively referred to as "the Plaintiffs”), initiated two groundbreaking lawsuits. They filed separate suits against Suno and Udio, two AI music-generating systems, in the United States District Court for the District of Massachusetts and the District Court for the Southern District of New York, respectively. These lawsuits mark the first legal actions against music-generating AI systems, which create music in response to a user’s text prompt and allow users to upload their audio files onto commercial streaming platforms such as Apple Music and Spotify. These cases are part of a recent trend of copyright owners seeking protections over the use and profit others make from their work.[1]

The Plaintiffs’ causes of action stem from the U.S. Copyright Act, specifically alleging copyright infringement in violation of 17 U.S.C. § 106 and copyright infringement of pre-1972 copyrighted recordings in violation of 17 USC §1401[2]. The complaints state that “there is nothing that exempts AI technology from copyright law or excuses AI companies from playing by the rules. This lawsuit seeks to enforce these basic principles.”[3]

In both complaints, Plaintiffs support their claims by providing examples of original copyrighted songs such as “Billie Jean” by Michael Jackson, “Johnny B. Goode” by Chuck Berry, and “The Thrill is Gone” by B.B. King, which they allege the AI music generators used to produce songs that “have a striking resemblance” to the songs owned by the record labels. Plaintiffs include as evidence of their allegations the transcriptions of select Suno & Udio outputs and the copyrighted recordings they resemble to illustrate their technical and musical similarities.[4]

Plaintiffs seek various forms of relief, including a declaration that Suno and Udio willfully infringed on the Plaintiffs’ protected sound recordings, injunctive relief restraining further infringement of the recordings, statutory damages for willful infringement under 17 USC § 504(c) for up to $150,000 per infringed work (662 songs for Suno and 1,670 songs for Udio[5]), as well as costs and attorneys’ fees and other relief that the court deems just and proper[6].

Some key cases against generative AI companies, such as Andersen v. Stability AI Ltd. and Kadrey v. Meta Platforms, Inc., set forth the standards used by courts when reviewing allegations of copyright infringement involving AI.[7] Kadrey v. Meta Platforms, Inc., widely referred to as the “Sarah Silverman Case” due to the involvement of the renowned comedian and author Sarah Silverman, centers on allegations of copyright infringement against Meta.[8] In the United States District Court for the Northern District of California, the plaintiffs accused Meta of unauthorized use of their copyrighted works to train Meta's LLaMA (Large Language Model Meta AI) language model. Following Meta’s motion to dismiss, the court stated that the plaintiffs did not provide specific examples or allegations showing substantial similarity between the outputs of the model and their copyrighted works. The court emphasized that to claim copyright infringement, there must be a demonstrable and substantial similarity between the copyrighted work and the allegedly infringing output. The plaintiffs failed to present concrete evidence or detailed comparisons supporting their claim that the outputs of the LLaMA models resembled their works in a way that constitutes infringement. Without such evidence, the court found no basis to uphold the plaintiffs’ claims of copyright infringement based on the model's outputs.[9] The court dismissed all claims at issue in the motion to dismiss with leave to amend only the copyright-related claims.

Similarly, in Andersen v. Stability AI Ltd., the plaintiffs commenced a class action lawsuit on behalf of themselves and other professional artists, alleging that Stability AI, DeviantArt, and Midjourney used their copyrighted artworks without permission to train an artificial intelligence model known as Stable Diffusion.[10] The plaintiffs claimed their works were included in the five billion images taken from the internet by Large-Scale Artificial Intelligence Open Network (LAION), which were used as training data for the AI model. As a result, the AI-generated images created by Stable Diffusion are allegedly derivatives of the training, infringing on plaintiffs’ copyrights. Following the defendants’ motions to dismiss, the court’s decision focused on the requirement for the plaintiffs to allege that the AI-generated works are substantially similar to their copyrighted works. The court found that the plaintiffs had failed to plausibly allege substantial similarity. Specifically, the court highlighted:

• The plaintiffs admitted that the output images from the Stable Diffusion model are unlikely to closely match any specific image in the training data, which weakens their claim of substantial similarity.[11]
• For a copyright infringement claim to proceed, the plaintiffs must provide detailed facts showing how the AI-generated images are substantially similar to their original works[12].

As a result, the court granted the motions to dismiss but allowed the plaintiffs the opportunity to amend their complaint. The plaintiffs were instructed to provide more specific allegations that clearly demonstrate how the AI-generated images resemble their copyrighted works.

Plaintiffs in both Kadrey and Andersen have since amended their complaints, providing more detailed allegations against the AI developers. While neither the Kadrey nor Andersen decisions defined “substantially similar” for purposes of copyright infringement, the Tenth Circuit has stated, in a copyright infringement case not related to AI, that: “Substantial similarity exists when ‘the accused work is so similar to the plaintiff's work that an ordinary reasonable person would conclude that the defendant unlawfully appropriated the plaintiff's [protectable] expression by taking material of substance and value.’”[13][10]

Based on the precedents set by Kadrey and Andersen, when alleging copyright infringement by AI, plaintiffs must demonstrate substantial similarity between the copyrighted work and the output generated by AI developers’ programs. Based upon these standards, by including examples of these similarities in their complaints, the Plaintiffs in the UMG actions may have successfully alleged copyright infringement claims against the two AI music-generating systems.

The cases commenced by the Plaintiffs in the UMG actions could help to clarify what is required for plaintiffs to successfully allege copyright infringement in the context of AI-generated content. The courts will now need to clarify how to apply this standard and determine what “substantially similar” means when applied in the context of AI-generated content.

Cullen and Dykman’s Intellectual Property team continues to monitor important developments in trademark and copyright law. Should you have any questions about this legal alert, please feel free to contact Karen Levin (klevin@cullenllp.com) at (516) 296-9110 or Ariel E. Ronneburger (aronneburger@cullenllp.com) at (516) 296-9182.  Thank you to Nnenna Onwuchekwa, a summer associate, who assisted in the preparation of this alert.

This advisory provides a brief overview of the most significant changes in the law and does not constitute legal advice. Nothing herein creates an attorney-client relationship between the sender and recipient.

Footnotes

[1] UMG Recordings, Inc. v. Suno, Inc., 1:24-cv-11611 (D. Mass.) UMG Recordings, Inc. v. Uncharted Labs, Inc., 1:24-cv-04777 (S.D.N.Y.).

[2] Complaint in UMG Recordings, Inc. v. Suno, Inc., 1:24-cv-11611, at ¶¶ 30-31; Complaint in UMG Recordings, Inc. v. Uncharted Labs, Inc., at ¶  37.

[3] Complaint in UMG Recordings, Inc. v. Suno, Inc., at ¶ 2; Complaint in UMG Recordings, Inc. v. Uncharted Labs, Inc., at ¶ 2.

[4] Complaint in UMG Recordings, Inc. v. Suno, Inc., at ¶ 54-67; Complaint in UMG Recordings, Inc. v. Uncharted Labs, Inc., at ¶ 55-69.

[5] Complaint in UMG Recordings, Inc. v. Suno, Inc, at Exhibit A; Complaint UMG Recordings, Inc. v. Uncharted Labs, Inc., at Exhibit A.

[6] Complaint in UMG Recordings, Inc. v. Suno, Inc. at pp. 32-33; Complaint in UMG Recordings, Inc. v. Uncharted Labs, Inc., at pp. 39-40.

[7] Andersen v. Stability AI Ltd., No. 23-CV-00201-WHO, 2023 U.S. Dist. LEXIS 194324 at *16 (N.D. Cal. Oct. 30, 2023); Kadrey v. Meta Platforms, Inc., No. 23-cv-03417-VC, 2023 U.S. Dist. LEXIS 207683 at *1 (N.D. Cal. Nov. 20, 2023).

[8] Kadrey v. Meta Platforms, Inc., No. 23-cv-03417-VC, 2023 U.S. Dist. LEXIS 207683 at *1

[9] Id. at *2

[10] Andersen v. Stability AI Ltd., No. 23-CV-00201-WHO, 2023 U.S. Dist. LEXIS 194324 at *16.

[11]  Id.

[12] Id. at *36.

[13] Craft Smith, LLC v. EC Design, LLC, 969 F.3d 1092, 1104 (10th Cir. 2020) (alteration in original) (internal citations omitted.