The United States District Court recently granted an interlocutory appeal in the case of Thomson Reuters v. ROSS Intelligence, which raises an important question on whether the use of copyrighted material to train artificial intelligence (AI) systems constitutes ‘fair use’ under U.S. copyright law. ROSS Intelligence sought the court’s permission for an appeal under 28 U.S.C. § 1292(b), a statute allowing appeals on interim legal rulings that may significantly affect the outcome of litigation. The court’s decision to permit this appeal signals the potential legal, ethical, and industry-wide ramifications surrounding AI and intellectual property law.
To provide some context, the issue stems from ROSS’s use of Westlaw’s headnotes and the West Key Number System, both proprietary tools by Thomson Reuters. Thomson Reuters alleges that ROSS violated the Copyright Act by using small excerpts of these materials—comprising 0.076% of the headnotes corpus—to train its AI-powered legal search engine. ROSS claims that this usage falls under ‘fair use,’ a legal doctrine codified under Section 107 of the U.S. Copyright Act of 1976, which permits limited use of copyrighted material for purposes like commentary, education, or research without the need for permission. However, earlier in this litigation, the court ruled that ROSS’s use was not transformative and did not meet the criteria for fair use.
By granting the interlocutory appeal, the court certified two crucial questions: (1) whether the Westlaw headnotes meet the originality requirement under copyright law, and (2) whether ROSS’s use of this content qualifies as transformative fair use. The originality question examines whether the headnotes involve a “creative spark” sufficient to be considered original works. Courts generally require more than mere factual compilations for works to qualify as original under copyright law, as established in landmark cases like Feist Publications, Inc. v. Rural Telephone Service Co. (499 U.S. 340, 1991). In this context, ROSS argues that the headnotes merely summarize factual statements, falling short of the creativity threshold.
Ethically, the case is pivotal because it ventures into the balance between incentivizing innovation through intellectual property and ensuring that legal and technological advancements are not stymied. Critics worry that a ruling against ROSS risks chilling AI innovation by creating excessive hurdles for training datasets. On the flip side, content creators argue that widespread unauthorized use of copyrighted material could undermine their economic interests and disincentivize original work production.
The ramifications for the AI industry are equally profound. Legal AI tools like ROSS hold the promise of democratizing access to complex legal knowledge by automating searches and reducing costs. If courts impose restrictive interpretations of copyright protection, developers might have fewer resources to train their algorithms, potentially stalling progress in machine learning and AI capabilities. Conversely, creators of proprietary data could gain more robust legal measures to protect their intellectual property, leading to increased litigation risks for AI developers reliant on third-party data.
The Third Circuit Court’s forthcoming decision on this interlocutory appeal could create precedential value in an area still relatively nascent in terms of legal scrutiny. Beyond impacting ROSS and Thomson Reuters, it will likely have sector-wide effects on how copyright law interacts with AI technology, potentially influencing future legislative frameworks or judicial interpretations globally.