ROSS: Opportunistic Infringement or Fair Use?

Before the release of ChatGPT in 2022, a start-up company, ROSS Intelligence Inc., founded by three students (two computer scientists and a lawyer) at the University of Toronto in 2014, attempted to create the first AI lawyer. Their effort began at the IBM Watson Cognitive Computing Competition with a prototype that trained Watson to read legal text and answer questions about case law. After winning second place at the competition, they continued working toward creating the first AI lawyer built on IBM Watson Natural Language Processing (NLP). They named their AI legal research system Ross. (Note this was an era pre-transformer AI architecture and pre-attention mechanism.)

During Ross’s development, they discovered Ross needed training data and unsuccessfully attempted to license the use of Westlaw headnotes from Thomson Reuters. Thomson Reuters declined, viewing Ross as a potential direct competitor. Undaunted, they forged ahead hiring a third party to create bulk memos using Westlaw headnotes to train Ross. Thomson Reuters responded by filing a lawsuit in May 2020 in the U.S. District Court, District of Delaware. After years of litigation and an interlocutory appeal, the U.S. Court of Appeals for the Third Circuit became the first federal appellate court to issue a fair use decision examining the use of stolen work, by the now-defunct AI startup ROSS Intelligence, Inc., to train Ross, its non-generative AI system. Thomson Reuters Enter. Ctr. GmbH v. ROSS Intel. Inc., No. 25-2153, __ F.4th __, 2026 WL 2932557 (3d Cir. Sept. 29, 2026). Affirming the District Court’s grant of partial summary judgment in favor of Thomson Reuters, the Third Circuit’s decision in ROSS signals how courts may analyze AI training and transformative use under 17 U.S.C. §107(1), distinguishing between training non-generative AI systems and generative-AI systems. This post examines “transformativeness” under Section 107(1) and its impact on the outcome in the ROSS litigation and subsequent interlocutory appeal to the Third Circuit.

Fair Use Doctrine and “Transformativeness”

“From the infancy of copyright protection, some opportunity for fair use of copyrighted materials has been thought necessary to fulfill copyright’s very purpose, ‘[t]o promote the Progress of Science and useful Arts….’ U.S. Const., Art. I, § 8, cl. 8.” Campbell v. Acuff-Rose Music, Inc., 510 U.S. 569, 575, 114 S. Ct. 1164, 1169, 127 L. Ed. 2d 500 (1994).

Fair use originated as a judge-made legal doctrine until it was later codified for the first time in the Copyright Act of 1976. In 1841, Justice Story explained in Folsom v. Marsh, that fair use analysis requires the court to “look to the nature and objects of the selections made, the quantity and value of the materials used, and the degree in which the use may prejudice the sale, or diminish the profits, or supersede the objects, of the original work.” Campbell, 510 U.S. 576 (quoting Folsom v. Marsh, 9 F. Cas. 342, 348 (No. 4,901) (CCD Mass. 1841)). These four factors are reflected in the current fair use doctrine under 17 U.S.C. § 107:

(1) the purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes;

(2) the nature of the copyrighted work;

(3) the amount and substantiality of the portion used in relation to the copyrighted work as a whole; and

(4) the effect of the use upon the potential market for or value of the copyrighted work.

The fair use factors “are to be explored, and the results weighed together, in light of the purposes of copyright.” Campbell, 510 U.S. at 578.

Notably, Section 107(1) does not include the term “transform.” Rather, Justice Souter, in Campbell v. Acuff-Rose Music, Inc., 510 U.S. 569 (1994), established the concept of “transformativeness,” an idea first articulated by Judge Pierre Leval in Toward a Fair Use Standard, 103 Harv. L. Rev. 1105, 1111 (1990), that asks under the first fair use factor, whether the secondary use displaces the original work or rather adds something new, with a “further purpose or different character, . . . in other words, whether and to what extent the new work is ‘transformative.’” Campbell, 510 U.S. 579. Since its adoption by the Court in 1994, transformative use has emerged as a dominant factor in fair use analysis, even if courts have widely diverged in its application. Jiarui Liu, An Empirical Study of Transformative Use in Copyright Law, 22 Stan. Tech. L. Rev. 163 (2019).

In 2023, the U.S. Supreme Court revisited the concept of “transformativeness” in Andy Warhol Found. for the Visual Arts, Inc. v. Goldsmith, 598 U.S. 508, 143 S. Ct. 1258, 215 L.Ed.2d 473 (2023). In Warhol, the Court explained that each specific use must be analyzed and that “[t]he same copying may be fair when used for one purpose but not another,” reaffirming transformative use under the first fair use factor, stating:

In sum, the first fair use factor considers whether the use of a copyrighted work has a further purpose or different character, which is a matter of degree, and the degree of difference must be balanced against the commercial nature of the use. If an original work and a secondary use share the same or highly similar purposes, and the secondary use is of a commercial nature, the first factor is likely to weigh against fair use, absent some other justification for copying.

Id. at 532-33.

Was ROSS’s Unauthorized Use Fair?

In 2020, Thomson Reuters filed a lawsuit against ROSS Intelligence Inc. alleging copyright infringement and tortious interference with contract. After years of litigation and an initial partial summary judgment opinion in 2023, Judge Stephanos Bibas, Circuit Judge, sitting by designation, invited the parties to revisit summary judgment before trial convened in August 2024. Reconsidering the issues, Judge Bibas revised parts of the 2023 partial summary judgment opinion, granting partial summary judgment to Thomson Reuters related to its copyright infringement claim and against ROSS’s fair use defense, and denying ROSS’s partial summary judgment motion as to these issues. Thomson Reuters Enter. Ctr. GMBH v. Ross Intel. Inc., 765 F. Supp. 3d 382, 390 (D. Del. Feb. 11, 2025), motion to certify appeal granted, No. 1:20-CV-613 (D. Del. May 23, 2025), and aff’d, No. 25-2153 (3d Cir. Sept. 29, 2026).

The District Court was not convinced by ROSS’s arguments that its alleged “intermediate copying” was in fact a transformative use of Thomson Reuters’s headnotes. Here, Judge Bibas distinguished a trio of “intermediate copying” cases: (1) Google LLC v. Oracle Am., Inc., 593 U.S. 1, 24, 141 S. Ct. 1183, 209 L.Ed.2d 311 (2021); (2) Sony Comput. Ent., Inc. v. Connectix Corp., 203 F.3d 596, 599, 606–07 (9th Cir. 2000); and (3) Sega Enters. Ltd. v. Accolade, Inc., 977 F.2d 1510, 1514–1515, 1522–23 (9th Cir. 1992), explaining that these cases concerned the copying of computer code that was necessary to reach the underlying ideas in order to innovate. Id. at 398-99. Instead, Judge Bibas applied the newer framework advanced by Warhol, examining the broad purpose and character of ROSS’s use of Thomson Reuters’s headnotes. Id. While its creators hoped Ross would be the first AI lawyer, Judge Bibas explained that they had built a competing legal research tool to Westlaw; the user enters a legal question, and Ross returns relevant judicial opinions (not new content like a generative AI system). Id. Essentially, Westlaw’s headnotes made it easier for ROSS to develop a competing legal research tool. Id. This proved fatal to ROSS’s transformative use argument.

The District Court found that ROSS’ use of Thomson Reuters’s headnotes was commercial and non-transformative. Id. at 397-99. The first fair use factor weighed against fair use. Id. Balancing the remaining fair use factors, Judge Bibas held that factors two and three favored fair use, but not factor four, and weighing the four factors together granted partial summary judgment for Thomson Reuters. Id. at 401.

The parties requested and the District Court certified an interlocutory appeal to answer two questions, one being: “whether Ross’s alleged use of the headnotes was fair use.” Thomson Reuters Enter. Ctr. GmbH v. Ross Intel. Inc., No. 1:20-CV-613-SB, 2025 WL 1488015 (D. Del. May 23, 2025).

The Third Circuit Rejects ROSS’s Fair Use Defense

On Sept. 29, 2026, the U.S. Court of Appeals for the Third Circuit affirmed the District Court, staking its claim as the first federal appellate decision to apply fair use under 17 U.S.C. § 107 to AI system training in Thomson Reuters Enter. Ctr. GmbH v. ROSS Intel. Inc., No. 25-2153, __ F.4th __, 2026 WL 2932557 (3d Cir. Sept. 29, 2026). On appeal, the Third Circuit explained: “In truth, this is no more than an ordinary copyright case,” ultimately rejecting ROSS’s fair use defense. Id. at *1, *10.

Examining the “purpose and character” of ROSS’s use under the first factor, the Third Circuit undertook the same analysis under Warhol as the District Court, examining if ROSS’s use was commercial and if its use was transformative, meaning does ROSS’s use have a purpose or character different from the original? Id. at *6-7. It was undisputed that ROSS’s use was commercial. Next, the Third Circuit considered whether ROSS’s use was transformative.

First, ROSS argued that its use of Thomson Reuters’s headnotes was analogous to Google’s scanning of entire books to create a search function with snippets of text, Authors Guild v. Google, Inc., 804 F.3d 202 (2d Cir. 2015). Id. at *7. The Third Circuit readily distinguished Authors Guild, explaining that Google’s use was highly transformative; the original purpose of the books was to be read, while Google’s purpose was to enable full-text searching, which may potentially lead users to purchase the original books. Id. Here, the court explained ROSS’s purpose mirrored Thomson Reuters’s original use of its headnotes; to help users find court opinions based on their search queries. Id. ROSS’s purpose, by its own admission, was to replace Westlaw. Id.

Second, ROSS argued that its use was similar to three “intermediate copying” cases, again relying on Oracle, Sony, and Sega. Id. The Third District rejected this argument, explaining that unlike the “intermediate copying” cases, “ROSS does not need to copy Thomson Reuters’s headnotes to access the underlying unprotected information. ROSS had access to the underlying judicial opinions and could freely copy them to make the memos needed to train its AI. It chose not to do so because copying the headnotes offered an ‘easy’ way to create its training memos . . . Unlike necessity, ease is not a justification for copying.” Id. (Emphasis in original).

Acknowledging ROSS’s intermediate step of using the headnotes to train its AI system “presents a slight degree of difference in use,” the Third Circuit explained “the undisputed evidence demonstrates that ROSS used the headnotes to train an AI program for the benefit of its legal-research platform.” Id. at *6. (Emphasis in original). Further explaining, “Thomson Reuters and ROSS use the headnotes to create and optimize a legal-research platform that helps users find responsive legal material,” the Third Circuit found ROSS shared “the same ultimate purpose as Thomson Reuters’s use, making ROSS’s use minimally transformative, at best.” Id. Adding in a footnote that ROSS’s “AI platform cannot generate original expression,” distinguishing this case from In re: OpenAI, Inc. Copyright Infringement Litig., No. 1:25-md-3143 (S.D.N.Y. Sept. 1, 2026) and Bartz v. Anthropic PBC, 787 F. Supp. 3d 1007 (N.D. Cal. 2025). Id. at n.7.

The Third Circuit found that ROSS’s use was “highly commercial and minimally transformative,” concluding that the first fair use factor weighed against fair use. Id. at *7. Like the trial court, the Third Circuit found the second factor weighed slightly in favor of fair use. Id. at *10. However, the court’s analysis of the third and fourth factors weighed against fair use. Id. “Evaluating these factors together, and in light of copyright’s purposes,” the Third Circuit held that “ROSS’s use was not fair.” Id.

ROSS Intends to Challenge the Third Circuit’s Decision

Was it easier to train Ross using Westlaw headnotes? Yes. But ROSS gambled and lost in the Third Circuit. The Third Circuit’s decision signals that non-generative AI systems, like Ross, likely do not satisfy the “transformativeness” requirement under the first fair use factor, which when weighing these factors together, in light of copyright’s purpose, will likely tip the balance against fair use.

However, ROSS is planning to petition the U.S. Supreme Court to review the Third Circuit’s decision. Yar Chaikovsky, partner at White & Case, issued this statement on behalf of ROSS:

“We respectfully disagree with the Third Circuit’s decision and believe it creates continued uncertainty around the application of copyright law to AI model training,” Chaikovsky said. “We intend to seek review by the Supreme Court to obtain much-needed clarity on these issues, which carry significant implications for innovation and the development of AI technologies.”

Stay tuned…this remains a developing area of the law with more questions than answers.