AI Training and Copyright Infringement: Lessons from the Ross Intelligence Case

Thomson Reuters sued Ross Intelligence for using its content to train its AI technology. Scott Hervey and Tara Sattler talk about this copyright dispute on this installment of The Briefing. Watch this episode on the Weintraub YouTube channel here.   Show Notes: Scott: Thompson Reuters, the provider of the Westlaw Legal Research Platform, sued Ross Intelligence for copyright infringement based on Ross's use of Westlaw content to train Ross's IP technology. This court ruling on a motion for summary judgment may provide guidance for future similar cases, and it even provides some additional guidance into the application of a post-Warhol fair use defense. We're going to talk about this case on the next installment of The Briefing by Weintraub Tobin. These are the basic facts underlying this lawsuit. Ross is an AI legal startup. Ross hired a subcontractor to create memos with legal questions and answers. The questions were meant to be those that a lawyer would ask, and the answers were direct quotations from legal opinions. Those memos were used to train Ross's AI tool. Thompson Reuters contends that these questions were essentially Westlaw case Headnotes. Ross denies that the Westlaw Headnotes were copied but also raises a fair use defense. As the case went forward, both sides moved for summary judgment on Ross's fair use defense. The court denied the party's motions for summary judgment on Ross's fair use defense, but there are a few points in this opinion that may shape the way future AI training cases play out. Tara: Before we get into the analysis of Ross's fair use defense, the court spent a significant amount of time talking about the scope of Westlaw's copyright. Westlaw's copyright extends to its Headnotes and its arrangement of the Headnotes and opinions, but its copyright does not extend to the opinions itself. Scott: That's right, and the reason the court spent so much time talking about the scope of Westlaw's copyright was because Ross challenged Westlaw's copyright claim in the Headnotes. Ross claims that the Westlaw Headnotes follow or closely mirror the language of the judicial opinions. And if a Headnote merely copies a judicial opinion, it's not copyrightable. But if it varies more than a trivial amount, then Westlaw owns a valid copyright. The court found that this leaves a genuine factual dispute about how original the Headnotes are. If the Headnotes are mere regurgitation of parts of an opinion, this will severely impact the strength and the extent of Westlaw's copyright case and Westlaw's copyright in its whole, including in the Headnotes. And it also goes to whether Ross was copying the Headnotes or the opinions themselves. Tara: So, how do you see this applying to other copyright cases involving AI training? Scott: Well, this analysis is part of the extrinsic test, which is used in the determination of substantial similarity after the plaintiff has identified specific criteria which it alleges have been copied. The court separates the unprotectable elements, such as facts or ideas, from the elements that are protectable. And then, it sorts out whether there is enough similarities between the works as to the elements that are protectable, such that a reasonable jury could find that the defendant's work is substantially similar to the protected elements of the plaintiff's work. This analysis is part of any copyright case, and it certainly will be part of an AI training case as well. Tara: Okay, so back to Ross's fair use defense, because the court finds that Ross actually copied the head notes. So fair use balances four factors: the purpose and character of the ...