This post was originally published on this site.
“Judge Bibas’ summary judgment rulings are doctrinally flawed… ROSS contends, because neither opinion answered the central question of factor one.”
Yesterday, artificial intelligence (AI) developer ROSS Intelligence filed a petition for the certification of an interlocutory appeal under 28 U.S.C. § 1292(b) to challenge originality and fair use findings entered in the District of Delaware in a copyright infringement case filed by global legal information company Thomson Reuters. ROSS Intelligence’s petition asks the U.S. Court of Appeals for the Third Circuit to reconsider the district court’s determinations that Westlaw’s headnotes and Key Number System are original, and that ROSS’ use of those materials to train its AI model for legal research wasn’t transformative.
Parroting Judicial Opinions Fails to Achieve Originality Required by Feist
Substantial grounds for a difference of opinion on the legal questions presented, a requirement for the certification of Section 1292(b) interlocutory appeals, is arguably made clear by conflicting opinions on controlling questions of law from U.S. Circuit Judge Stephanos Bibas, a Third Circuit judge sitting by designation due to Judge Leonard Stark’s confirmation to the Federal Circuit. In September 2023, Judge Bibas entered an opinion largely denying motions for summary judgement from both ROSS and Thomson Reuters, finding that both the originality and fair use issues should move toward trial. However, this February, Judge Bibas issued a revised ruling finding that Westlaw’s headnotes met the low threshold for originality, while also finding no fair use by ROSS because it had developed a competing legal research product.
Noting the standard on the originality required for copyright protection from Feist Publications v. Rural Telephone Service Company (1991), in which the Supreme Court wrote that “facts themselves do not become original through association” with original compilations of those facts, ROSS argues that both of Judge Bibas’ summary judgment rulings erred finding any potential copyrightability for Westlaw’s headnotes, “which parrot [uncopyrightable] judicial opinions as best they can.”
ROSS contends that Westlaw’s headnotes share the same defects in originality that befell Westlaw’s case reports, which failed the originality standard in the Second Circuit’s 1998 decision in Matthew Bender & Co. v. West Publishing Co. In that case, the appellate court found that the case reports “lack[ed] a modicum of creativity,” only providing the kind of “garden variety” information that failed to result in copyright protection for the telephone directory at issue in Feist. ROSS contends that Westlaw’s headnotes are similarly dictated by industry conventions, and that editorial directions to follow the language of court opinions as closely as possible meant that it was legal error to consider those summaries original.
ROSS Asks Third Circuit to Fix Judge Bibas’ Rigid Application of Fair Use Test
Although the fair use framework codified by Congress at 17 U.S.C. § 107 has inherent ambiguities, ROSS’ petition argues that those doctrinal features reflect the judge-made origins of fair use. Still, both of Judge Bibas’ summary judgment rulings are doctrinally flawed regarding their application of the four-factor fair use test, ROSS contends, because neither opinion answered the central question of factor one, which looks at the purpose and the character of the use.
Judge Bibas’ first ruling found that a jury must decide whether ROSS engaged in “transformative intermediate copying” by translating the human language of Westlaw headnotes into a computer-understandable format in the creation of ROSS’ search tool that produces highly relevant quotations from judicial opinions in response to human language questions. In his second ruling, Judge Bibas found it impossible that ROSS could engage in transformative intermediate copying because it developed a product competing with Thomson Reuters. In neither decision did Judge Bibas address whether ROSS’ AI-assisted legal research tool added something new with a different character than Thomson Reuter’s manual headnote sorting system.
As to the fourth fair use factor, the impact of the use on the potential market for the work, ROSS similarly argues legal errors in both of Judge Bibas’ rulings. In acknowledging that ROSS’ use might be transformative, Judge Bibas’ first summary judgment ruling recognized that its AI tool might serve a different purpose such that it’s not a market substitute for Westlaw. While this reasoning was jettisoned from Judge Bibas’ second ruling, both rulings arguably erred in basing the ultimate finding of factor four on impacts to hypothetical markets that don’t exist, here the potential market for AI training data. “[E]qually jaw dropping” was Judge Bibas’ ultimate determination that the “public has no right to Thomson Reuters’s parsing of the law,” with ROSS pointing out that this assertion is irrelevant because ROSS did not enable public access of Westlaw headnotes.
Finally, ROSS challenges Judge Bibas’ balancing of the fair use factors, especially his arguably unfounded determination that factors one and four weigh most heavily in the analysis. This rigid application of the fair use test led Bibas in both rulings to ignore that ROSS only used 0.076% of Westlaw’s total headnotes, which speaks to the third factor’s assessment of the amount and substantiality of the portion used. Although Judge Bibas found factor three weighing in ROSS’ favor, the little weight afforded this factor in light of ROSS’ de minimis use “threw the court’s balancing into disarray,” according to the petition.
ROSS’ petition concludes by noting that the Third Circuit’s determinations on these controlling questions of law will materially advance the litigation by resolving legal issues instead of submitting them to a costly jury trial. Further, the appellate court’s ruling would serve the public benefit by answering important questions of copyright law in the AI context, which could impact national security given the global race toward AI dominance.
Image Source: Deposit Photos
Author: Tolikoff
Image ID: 187828250