Court Grants Summary Judgment in AI Copyright Clash, Rejecting "Fair Use"
In Short
The Situation: There are numerous actions currently pending in federal courts that involve artificial intelligence ("AI") and copyright law. Among other things, courts must determine if infringement occurs when an AI is trained on copyrighted materials without a license from the owner. And, if copyright infringement occurred, courts must then address whether training an AI with copyrighted materials constitutes fair use.
The Development: In February 2025, a Delaware federal judge granted summary judgment to Thomson Reuters, ruling that the use of Westlaw Headnotes to train a competing legal research AI tool is not fair use. This marks the first substantive ruling on direct copyright infringement and fair use in the context of AI.
Looking Ahead: Although this decision provides some insights on how one court has grappled with fair use in the AI context, this defense remains fact-specific. Stakeholders should monitor further guidance on generative AI, and other issues involving training AI with copyrighted materials—which will emerge in litigation and in the Copyright Office's forthcoming third part of its ongoing AI Report.
On February 11, 2025, a copyright owner prevailed in a hard-fought battle involving AI and copyright, as a federal district judge granted partial summary judgment in favor of Thomson Reuters in its longstanding litigation against Ross Intelligence ("Ross"). In particular, the court found Ross's use of Westlaw Headnotes to train its AI legal research tool constituted direct copyright infringement and rejected all of Ross's defenses, including fair use. The judge revised his earlier order in which he denied the parties' cross motions for summary judgment. While issues of infringement and fair use are notoriously fact-specific, this ruling provides insight into how courts and litigants may address critical issues arising at the intersection of AI and copyright.
In assessing infringement, the court emphasized the "extremely low" threshold for originality set out in the Supreme Court's seminal decision in Feist Publications, Inc. v. Rural Telephone Service Co., 499 U.S. 340, 345 (1991), and concluded that certain Thomson Reuters's Westlaw Headnotes qualify for copyright protection—thus rejecting Ross's argument that these Headnotes were uncopyrightable summaries of public judicial opinions. Assessing the evidence (including expert opinions), the court granted summary judgment of direct copyright infringement by Ross, finding that Ross had actually copied protected material from the Headnotes to train and/or validate its AI legal research tool.
As to Ross's asserted defenses, the court declared: "[n]one of Ross's possible defenses holds water. I reject them all." This included Ross's innocent infringement, copyright misuse, and merger defenses, as well as Ross's asserted fair use defense. The court devoted the most attention to fair use—a defense that is central to numerous pending copyright infringement litigations involving AI. Noting that fair use is a mixed question of law and fact, the court concluded that the "undisputed facts here push this case squarely into the legal realm." Thomson Reuters Enter. Ctr. GMBH v. Ross Intel. Inc., No. 1:20-CV-613-SB, at 14, 16 (D. Del. Feb. 11, 2025). The court analyzed the statutory factors for fair use as follows:
- Purpose of Use: The court reasoned that Ross's tool directly competes with Westlaw, constituting commercial use, and Ross's use of the Headnotes is not transformative because Ross's AI tool "does not have a 'further purpose or different character'" from Thomson Reuters's. Id. at 17 (citing Andy Warhol Found. for the Visual Arts, Inc. v. Goldsmith, 598 U.S. 508, 529 (2023)). Accordingly, the court found that the first factor heavily favored Thomson Reuters. Importantly here, the court emphasized that Ross's AI "is not generative AI (AI that writes new content itself)." Id. As numerous pending disputes do involve generative AI tools, it is likely that defendants in these cases will emphasize that generative AI creates new content and is thus "transformative." Indeed, the court reiterated this distinction later in its opinion: "Because the AI landscape is changing rapidly, I note for readers that only non-generative AI is before me today." Id. at 19.
- Nature of the Works: The court found that the Headnotes have "more than the minimal spark of originality required for copyright validity. But the material is not that creative." Id. at 20. This factor therefore favored Ross.
- Portion of Works Copied: Here the court found that there was no dispute that "Ross's output to an end user does not include a West headnote" and emphasized that "[w]hat matters" is the amount and substantiality of what is made accessible to the public, rather than the amount and substantiality used in making a copy. Id. at 21. The court concluded that this factor favored Ross.
- Market Impact: Emphasizing that this factor "is undoubtedly the single most important element of fair use[,]" the court found that this factor favored Thomson Reuters. Here the court recognized not just the original market for legal-research platforms, but also "at least one potential derivative market" for AI training data. As Ross has the burden of proof on fair use, the court found that it had "not put forward enough facts to show that these markets do not exist and would not be affected." Id. at 21–22.
As the first time a court has made a dispositive ruling on fair use in the context of AI tools, this decision sheds light on how federal judges may approach the fair use question in copyright litigation involving AI technology. Stakeholders should watch for additional analyses to emerge from other pending cases involving AI and copyright, as well as forthcoming guidance from the Copyright Office that will focus on the legal implications of training AI models on copyrighted works.
Three Key Takeaways
- Purpose and Character of the Use Focuses on Commerciality and Transformativeness: Factor one of the fair use analysis addresses the purpose and character of the defendant's use which, as the court noted here, centers on whether the use was commercial and whether it was transformative. Commercial uses by a competitor are less likely to be fair and whether a work is deemed "transformative" will be fact-specific.
- Market Impact is Critical—Both Actual and Potential Markets: The court emphasized that market impact is the most important factor in assessing fair use, and considered the effect of copying on the market for the original work as well as the potential derivative market for data to train AIs.
- Case-by-Case Determinations: While the court emphasized that the case before it did not involve generative AI, it does not follow that all generative AI uses will necessarily be seen as "transformative." Ultimately, the fair use analysis remains fact- and context-specific, and insights on what is "fair" in the context of non-generative AI and generative AI will continue to play out in litigation.
