Existing precedent goes neg.

Edward Lanquist and Jeremy Ray, “Artificial Intelligence and Copyright Law: The NYT v. OpenAI – Fair Use Implications of Generative AI,” Baker Donelson, 02-05-2024, https://www.bakerdonelson.com/artificial-intelligence-and-copyright-law-the-nyt-v-openai-fair-use-implications-of-generative-ai

Relevant case law from the Second Circuit, Ninth Circuit, and United States Supreme Court helps guide the fair use analysis. The below-referenced opinions support a likely finding of fair use as it relates to using unlicensed content to simply train LLMs.

In 2015, the Second Circuit found that Google's unauthorized digitizing of copyright-protected works, creation of a search functionality, and display of snippets of those works were non-infringing fair uses. Authors Guild v. Google, Inc., 804 F.3d 202, 229 (2d. Cir. 2015). The Second Circuit reasoned that the purpose of the copying was highly transformative, the public display of the text was limited, and the revelations did not provide a significant market substitute for the protective aspects of the originals. Id. The Second Circuit found that the fact that Google's use was of a commercial nature and had a profit motivation, did not justify denial of fair use. Id. In Sega Enterprises, Ltd. v. Accolade, Inc., 977 F.2d 1510 (9th Cir. 1992), the Ninth Circuit found copying software code to permit a video game to run on a console to be fair use. Likewise, in Perfect 10, Inc. v. Google, Inc., 508 F.3d 1146 (9th Cir. 2007), the Ninth Circuit found a search engine's collection and display of thumbnail images to be a fair use.

In Google LLC v. Oracle America, Inc., 141 S. Ct. 1183 (2021), the Supreme Court found that Google’s copying of Oracle's Java SE API was a fair use of such material. The Supreme Court focused its fair use analysis on whether the use was transformative (i.e., whether it adds something new, with a further purpose or different character). Id. at 1202-1203. The opinion noted that Google copied the API only insofar as needed to include tasks that would be useful in smartphone programs and only insofar as needed to allow programmers to call upon those tasks without discarding a portion of a familiar programming language and learning a new one, which supported a finding of fair use. Id. at 1203. These facts supported the fair use finding.

The Supreme Court's recent Andy Warhol Found. for the Visual Arts, Inc. v. Goldsmith, 598 U.S. 508 (2023), opinion similarly focused on whether the uses at issue were transformative. In Goldsmith, the Supreme Court found that, while the use of a copyrighted work may be fair if the use has a purpose and character that is sufficiently distinct from the original, the uses at issue before the Court were not transformative because they shared substantially the same commercial purpose (i.e., to illustrate a magazine about Prince with a portrait of Prince). Id. at 541-550.