The recent decision by the District Court to grant an interlocutory appeal in the case of Thomson Reuters v. Ross Intelligence has significant implications for artificial intelligence (AI) and intellectual property law. The court, citing 28 U.S.C. § 1292(b), certified two key questions for appeal: (1) whether the Westlaw headnotes and West Key Number System qualify as original works under the Copyright Act, and (2) whether the use of a fractional percentage (0.076%) of these headnotes by Ross Intelligence constitutes fair use, particularly in the context of AI training purposes. This proceeding is poised to influence the evolving intersection of copyright law and AI technology.
The legal crux of this case involves the interpretation of two elements central to the Copyright Act of 1976—the originality requirement and the four-factor fair use test. Originality, as established in Feist Publications, Inc. v. Rural Telephone Service Co., requires that a work must exhibit a “modicum of creativity.” The first question certified for appeal probes whether Westlaw’s headnotes, which summarize legal rulings, meet this creativity threshold, or whether they are merely factual and devoid of a “creative spark.” Ross Intelligence has argued that deeming such works original could expand copyright protection to factual content, restricting innovation and setting a problematic precedent for machine learning applications that rely on vast datasets.
The second question, focusing on fair use, raises significant ethical and industry-specific concerns. Ross’s use of Westlaw’s headnotes involves tiny data fragments for the purpose of developing an AI-driven legal search platform. Under the four-factor fair use analysis, Ross contends that their use is transformative as it enhances legal research capabilities through AI, without supplanting Westlaw’s market value. They also highlight the negligible proportion of content used, approximately 0.076%, as evidence of minimal impact. If appellate courts find in favor of Ross on this issue, it could mitigate fears of a chilling effect on AI innovations, where copyright overreach might stifle access to raw data essential for training algorithms.
This case is emblematic of broader tensions in the industry regarding intellectual property rights in the age of AI. Companies developing AI systems often need to access and process protected materials to build functional models, raising questions about the balance between fostering innovation and protecting intellectual property. For example, content scraping to train AI tools has already led to disputes, such as those involving OpenAI, where AI models trained on publicly accessible internet data have been challenged for allegedly infringing copyrights.
The outcome of this interlocutory appeal will likely set critical legal precedents for the scope of fair use in AI development. It is also expected to influence policymaking efforts, such as recent initiatives by the U.S. Copyright Office and global regulators, to address gaps in existing intellectual property laws as they relate to machine learning technologies. Regardless of the appellate court’s decision, this case underscores the urgent need for clear guidelines that harmonize copyright protections with innovation in artificial intelligence, benefiting industries and society alike.