The US Court of Appeals for the Third Circuit has issued a landmark ruling finding Ross Intelligence liable for copyright infringement against Thomson Reuters. The lawsuit was triggered by the use of annotations from the authoritative Westlaw legal system to train Ross's own AI-powered legal search platform. This incident establishes an important legal precedent in the fields of intellectual property protection and machine learning regulation.
Core Conflict and Arguments of the Parties
Ross developed an innovative AI-based legal search platform designed to help lawyers find relevant excerpts from court rulings. To properly train the neural network, developers utilized Westlaw's editorial annotations, which feature concise summaries and legal highlights of court decisions. In response to the allegations, Ross argued that these materials were not eligible for copyright protection due to their close connection with the underlying judicial texts, and insisted their use was fair and transformative for educational purposes.
Court's Legal Evaluation and Fair Use Doctrine
The judicial instance rejected the defendant's arguments, ruling that Westlaw annotations meet the required minimum threshold of originality thanks to meticulous editorial effort. The court emphasized that using protected data to build a competing commercial service within the same market segment cannot be considered fully transformative. Furthermore, judges pointed to direct economic harm to the copyright holder and the creation of barriers to the emerging market of licensing content for AI training.
Implications for the Artificial Intelligence Industry
The issued verdict demonstrates a fundamental shift in the judicial system's approach to disputes surrounding artificial intelligence and copyright law. Technology developers can no longer rely on the argument that utilizing proprietary data at intermediate stages of AI training automatically exempts them from liability. The outcome clearly shows that algorithm creators must strictly consider the origin and status of the information arrays they employ.
