Generative Artificial Intelligence (AI) tools—defined by the U.S. Copyright Office as  technology that is “capable of producing outputs such as text, images, video, or audio (including emulating a human voice) that would be considered copyrightable if created by a human author”[1]—are trained by analyzing vast amounts of data, which may (and often do) include copyrighted works.[2]  With its enormous potential and untold consequences, generative AI has rapidly become a critical topic for creatives across disciplines.  For instance, just in the past year, among other controversies, generative AI played a significant role in strikes by writers and actors in Hollywood[3] and led to a dispute in which one of the world’s biggest music catalogues temporarily pulled its songs from TikTok,[4] while photographer Ansel Adams’s estate resorted to social media pressure to remove from Adobe’s stock photos AI-generated images resembling Adams’s work.[5]  Generative AI also is a continuing hot topic in media, with news companies arguing that generative AI plagiarizes journalists’ work.[6] 

The U.S. Copyright Office is in the midst of an ambitious investigation into the intersection of generative AI and copyright law and policy, which it launched in March 2023 with a series of listening sessions with various stakeholders.  Recordings and transcripts of the listening sessions, covering literary works, the visual arts, audiovisual works and music and sound recordings, are available on the Copyright Office’s website.  In June 2023, the Copyright Office held a webinar addressing basic questions about registration of works containing AI-generated materials, noting the importance of disclaiming purely AI-generated materials that are not eligible for copyright.  Then, in August 2023, it issued a Notice of Inquiry in the Federal Register on generative AI,[7] outlining questions related to four topics:

  1. Use of copyrighted works to train generative AI models—how datasets are selected or used to train generative AI, and what kinds of permission and compensation rights should be afforded to copyright holders.[8]
  2. Copyrightability of materials generated using AI systems—although “the law is clear that copyright protection in the United States is limited to works of human authorship,” where to draw the line between human-authored and AI-authored material.[9]
  3. Potential liability for infringing AI-generated works—how to apportion liability for infringing AI-generated works, for example, as between an AI system’s designer and its user where an AI work generated on that system infringes on a copyright.[10]
  4. Treatment of generative AI outputs that imitate the identity or style of human artists—how to address generative AI’s ability to “mimic [artists’ and performers’] voices, likenesses, or styles,” which typically are not protected by copyright law, but may implicate other legal issues, like the right to publicity or unfair competition.[11]

The Copyright Office received more than 10,300 written comments, submitted by a wide range of stakeholders, in response to its questions on these topics.[12]  For example:

  • The News/Media Alliance, a prominent industry association representing news and media publishers in the United States, submitted a legal opinion that use of copyrighted works to train generative AI is not “fair use.”[13]  Digital Content Next, a trade organization representing digital content companies, also analyzed the fair use doctrine, and asked the Copyright Office to “make clear that use of copyrighted works to train AI is not per se a fair use.”[14]
  • The McKinley Park News, a news company local to Chicago’s McKinley Park neighborhood, described vividly how its content had been “hoovered en mass[e]” into AI models, which used its content to keep users on other platforms, and warned that it would be forced to stop publishing if the current situation continued.[15]
  • The Science Fiction and Fantasy Writers Association raised concerns about the overwhelming quantity of AI-generated submissions to science fiction publications and lamented an imagined future without new human-authored works.[16] 
  • Scores of artists and other interested individuals voiced concerns that generative AI will prevent artists’ fair renumeration for their work, and squelch students’ opportunities to develop important skills.[17]
  • Universal Music Group sounded alarm bells about dangers to the music publishing industry of training AI with copyrighted material, but also recognized the promise of generative AI for creative, productive and non-infringing ends.[18] 
  • Other commenters noted that generative AI could “allow increased access to the creative arts for individuals with disabilities, such as long covid and traumatic brain injuries.”[19] 
  • Technology companies—including Meta, Apple and Google—submitted comments defending and championing their work in developing generative AI.[20]  For example, Meta argued that use of copyrighted material to train generative AI tools does not trigger copyright laws or, in the alternative, is fair use[21]; OpenAI argued that new technologies will democratize creativity and otherwise benefit science, productivity and economic growth[22]; and Apple argued that generative AI could aid in the development of computer code.[23] 
  • Microsoft expressed an openness to supporting some level of regulation, such as “new [labeling] requirements as they emerge” and new federal law to protect individuals against unauthorized use of name, likeness, image or signature.[24] 

Meanwhile, both the Copyright Office Review Board and U.S. courts have considered copyright applications for AI-generated work.  The Review Board has upheld copyright registration denials for works, or portions of works, involving generative AI, on grounds of insufficient human authorship where the purported copyright holder fed instructions or specific input into AI applications.[25]  In August 2023, the District Court for the District of Columbia upheld the Copyright Office’s decision that no copyright existed in a work generated entirely by AI, noting that “courts have uniformly declined to recognize copyright in works created absent any human involvement.”[26]  The D.C. District Court’s decision is currently on appeal, with oral argument scheduled for September 2024.[27]  There are also several pending private suits against AI companies for copyright violations in their platform development.[28]

The Copyright Office intends to release a report in several sections on its AI initiative in 2024, the first of which “will focus on digital replicas,” and the second of which “will address the copyrightability of works incorporating AI-generated material.”[29]  Subsequent reports “will focus on the topic of training AI models on copyrighted works as well as any licensing considerations and liability issues.” [30]  In addition, the Copyright Office is updating the Compendium of U.S. Copyright Office Practices to address treatment of AI.[31]  We will continue to monitor these important developments.

[1] See 88 Fed. Reg. 59,942 (Aug. 30, 2023). 

[2] See id at 59,943 n.2.  For a more in-depth explanation, the Copyright Office refers to an IBM research blog.  See Kim Martineau, What is generative AI?, IBM Research Blog (Apr. 20, 2023), https://research.ibm.com/blog/what-is-generative-AI.

[3] Joseph Gordon-Levitt, Opinion: If artificial intelligence uses your work it should pay you, Washington Post (July 26, 2023), https://www.washingtonpost.com/opinions/2023/07/26/joseph-gordon-levitt-artificial-intelligence-residuals/.

[4] Kristin Robinson, UMG and TikTok Strike Licensing Deal After Three-Month Standoff, Billboard (May 2, 2024), https://www.billboard.com/business/tech/umg-tiktok-strike-licensing-deal-1235671492/.

[5] Alex Greenberger, Ansel Adams Estate Calls Out Adobe for Selling AI-Generated Art Using Photographer’s Name, ARTNews (June 3, 2024), https://www.artnews.com/art-news/news/ansel-adams-estate-ai-generated-images-adobe-controversy-1234708739/.

[6] See John Herrman, The Other Big Problem With AI Search, New York Magazine (June 14, 2024), https://nymag.com/intelligencer/article/perplexity-demonstrates-the-other-big-problem-with-ai-search.html.

[7] See 88 Fed. Reg. 59,942 (Aug. 30, 2023).

[8] Id. at 59,945.

[9] Id.

[10] Id.

[11] Id.  For example, OpenAI delayed use of one of its ChatGPT “voices” after actress Scarlett Johansson alleged that it sounded strikingly like her voice.  See Tripp Mickle, Scarlett Johansson Said No, but OpenAI’s Virtual Assistant Sounds Just Like Her, N.Y. Times (May 20, 2024), https://www.nytimes.com/2024/05/20/technology/scarlett-johannson-openai-voice.html.  At the time, the general counsel to SAG-AFTRA, the actor’s union, warned that it was prepared to “aggressively” educate the technology industry that there are rights to a voice.  See Dan Milmo, Scarlett Johansson’s OpenAI clash is just the start of legal wrangles over artificial intelligence, Guardian (May 27, 2024), https://www.theguardian.com/technology/article/2024/may/27/scarlett-johansson-openai-legal-artificial-intelligence-chatgpt.

[12] See Regulations.gov, Artificial Intelligence and Copyright, https://www.regulations.gov/docket/COLC-2023-0006/comments.

[13] Comments of the News/Media Alliance, Artificial Intelligence and Copyright, Docket No. 2023-6 (Oct. 30, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-8956.

[14] Comments of Digital Content Next (Oct. 30, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-8736.

[15] Justin Kerr, Comments of the McKinley Park News (Nov. 10, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-10011.

[16] Comments of the Science Fiction Writers Association (Oct. 30, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-8834.

[17] See, e.g., Comment from Emily Baker (Sept. 6, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-9597 (“I am an author and artist and have quite a few artists in my immediate circle, and we are all concerned about the way AI is being used in art. . . . Allowing generative AI to train on copyrighted works without gaining the creator’s consent violates their right to control how their work is used, and makes it harder to receive fair compensation for their skills.”); Comment from Anonymous (Sept. 6, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-9600 (“i am a concerned youth and aspiring artist in many avenues of creativity and i strongly feel that AI works SHOULD NOT BE ALLOWED TO BE COPYRIGHTED.”); Comment from Alexandra Jocic (Sept. 26, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-9711 (“As an educator at a post-secondary institution, I have seen how AI is to the detriment of students who, given the opportunity, will forego [sic.] learning important skills in favor of using AI.”).

[18] See Comments of Universal Music Group (Oct. 30, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-9014.

[19] See, e.g., Comment from Anonymous (Sept. 6, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-9617.

[20] See Comment of Meta Platforms, Inc. (Oct. 30, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-9027; Comment of Apple Inc. (Oct. 30, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-9022; Comment of Google (Oct. 30, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-9003.

[21] Comment of Meta Platforms, Inc. (Oct. 30, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-9027

[22] See Comment of OpenAI (Oct. 30, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-8906.

[23] See Comment of Apple Inc. (Oct. 30, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-9022.

[24] See Comment of Microsoft Corporation (Oct. 30, 2023), available online at https://www.regulations.gov/comment/COLC-2023-0006-8750.

[25] See, e.g., Review Board Letter to Alex P. Garens, Esq. re: Second Request for Reconsideration for Refusal to Register SURYAST (SR # 1-11016599571; Correspondence ID: 1-5PR2XKJ) (Dec. 11, 2023), https://www.copyright.gov/rulings-filings/review-board/docs/SURYAST.pdf (no copyrightable material where applicant fed a picture he took into software and selected style); Review Board Letter to Tamara Pester, Esq. re: Second Request for Reconsideration for Refusal to Register Théâtre D’opéra Spatial (SR # 1-11743923581; Correspondence ID: 1-5T5320R) (Sept. 5, 2023), https://www.copyright.gov/rulings-filings/review-board/docs/Theatre-Dopera-Spatial.pdf (no copyright can be issued unless author is able to disclose which portion of the work was AI-generated material); Review Board Letter to V. Lindberg re: Zarya of the Dawn (Registration # VAu001480196) (Feb. 21, 2023), https://www.copyright.gov/docs/zarya-of-the-dawn.pdf (copyright certificate could only cover human-generated expressive material, not AI-generated material where prompts led to unpredictable output).

[26] See Thaler v. Perlmutter, 2023 U.S.P.Q.2d 980, at *6 (D.D.C. 2023). 

[27] See Order, Thaler v. Perlmutter, No. 23-05233 (Dkt. No. BL-25) (D.C. Cir. June 17, 2024).

[28] See, e.g., Alter v. OpenAI, No. 1:23-cv-08292-SHS (S.D.N.Y.) (proposed class action alleging that OpenAI and Microsoft use of materials for training generative AI a copyright infringement; related cases involve allegations of copyright violations by a nonprofit newsroom and the New York Times); In re OpenAI ChatGPT Litig., No. 3:23-cv-03223-AMO (N.D. Cal.) (consolidated action by book authors against OpenAI alleging copyright violations in training ChatGPT).

[29] See Nora Scheland, Looking Forward: The U.S. Copyright Office’s AI Initiative in 2024, Copyright Office (Mar. 26, 2024), https://blogs.loc.gov/copyright/2024/03/looking-forward-the-u-s-copyright-offices-ai-initiative-in-2024/.

[30] See id.

[31] See id.