OpenAI’s Sora & the Role of the US Copyright Office
By Bart Mueller
When Shira Perlmutter took office as the 14th Register of Copyrights in October 2020, many saw the US Copyright Office as relatively unimportant—a group of glorified librarians tucked away in a sleepy back room of the Library of Congress.[1] Four years later, this could not be farther from the truth. On February 15, 2024, OpenAI announced “Sora,” their new generative text-to-video Artificial Intelligence (AI) model, capable of outputting hyper realistic videos of up to a minute in length.[2] Prompts as simple as “A cartoon kangaroo disco dances” or “Historical footage of California during the gold rush,” will output content nearly indistinguishable from that created by the most skilled humans.[3] Understandably, many are calling it the “end of Hollywood.”[4] Though Sora will likely not be made available to the general public anytime soon,[5] its announcement brings the Copyright Office into the forefront of a national debate and perhaps demonstrates the need for it to transform from a quiet custodian of records and step up to its role as a central figure in the evolving narrative of intellectual property law.
Much of the legal debate around of Sora, and more broadly generative AI, can be traced to the question of authorship. Embedded in the fabric and constitutional basis of our copyright laws is the idea that copyright serves to protect “authors” so as to promote creativity. Given that it is unlikely that the framers had Sam Altman in mind at the constitutional convention, it is implicit that only human authors may be receive copyright in their creations.[6] Unsurprisingly, the 9th Circuit agreed with this assumption in rejecting PETA’s claim that a monkey had a copyright in a selfie it took on accident (presumedly they advocated for him to receive his royalties in Bananas).[7] And such, before OpenAI came on the scene, the law was relatively settled: authorship requires humanity.
Yet, enter Stephen Thaler, who, in a lawsuit against Perlmutter and the Copyright Office for denying his copyright in a work he specifically disclosed as generated by AI, contended that artificial intelligence could fulfill the role of an author.[8] For countless reasons, however, giving AI a copyright is a terrible idea. AI lacks the creative capacity, human authorship, originality, and creative choices required for copyright protection.[9] The primary goal of copyright law is to incentivize human innovation, not that of an algorithmic machine. Thankfully, the copyright office agreed and in a March 2023 statement of policy, announced that “AI-generated content that is more than de minimis should be explicitly excluded from the [copyright] application.”[10] They further advised that applicants have a duty to disclose any use of AI, based on it having “bearing upon the preparation or identification of the work or the existence, ownership, or duration of the copyright.”[11]
To those in the Intellectual Property (specifically trademark) practice, this disclosure requirement may seem familiar. Trademark disclaimers involve voluntarily relinquishing the exclusive right to use certain generic or descriptive elements of a trademark, where either the applicant voluntarily does so, or an examining attorney will require it before registration.[12]. Perhaps this parallel suggests one path forward: a well-staffed Copyright Office of examining attorneys, using AI detection software and other evaluation methods to evaluate whether a work was created significant amounts of AI. Whether or not this is the appropriate solution or direction is still unclear. Yet, the emergence of AI technologies like Sora has undoubtedly catalyzed a significant shift in the importance Copyright Office, putting them at the helm of the sinking ship that is our copyright laws in the age of AI.
Bart Mueller is a current 2L at Vanderbilt Law School. Prior to law school he attended the University of Georgia where he majored in Political Science and International Affairs.
[1] Cecilia Kang, The Sleepy Copyright Office in the Middle of a High-Stakes Clash Over A.I., N.Y. Times, (Jan. 25, 2024), https://www.nytimes.com/2024/01/25/technology/ai-copyright-office-law.html.
[2] Sora, OpenAI, https://openai.com/sora (last visited Feb. 27, 2024).
[3] OpenAI, Introducing Sora — OpenAI’s text-to-video model, YouTube (Feb. 16, 2024), https://www.youtube.com/watch?v=HK6y8DAPN_0&t=5s.
[4] Angela Yang & Brian Cheung, Hollywood is not ‘over’ just because of OpenAI’s Sora, some filmmakers say, NBC News (Feb. 22, 2024, 11:43 AM), https://www.nbcnews.com/tech/openai-sora-filmmakers-reaction-rcna139646.Top of Form
[5] Open AI has made it clear they will be withholding Sora from the general public indefinitely, so they have a chance to consult with policymakers, educators, and artists about its best use. OpenAI, supra note 1. Top of Form
[6] U.S. Const. art. I, § 8, cl. 8.
[7] Naruto v. Slater, 888 F.3d 418 (9th Cir. 2018).
[8] Thaler v. Perlmutter, No. CV 22-1564 (BAH), 2023 WL 5333236 (D.D.C. Aug. 18, 2023).
[9] Daniel J. Gervais, The Machine as Author, 105 Iowa L. Rev. 2053 (2020).
[10] Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence, 88 Fed. Reg. 16190-01.
[11] 17 U.S.C. § 409(10).
[12] 3 McCarthy on Trademarks and Unfair Competition § 19:63 (5th ed.).