Judge: Nvidia's Shadow Library Scripts 'Have No Other Purpose' Than Infringement (torrentfreak.com)

NVIDIA faces a significant legal battle as U.S. District Judge Jon Tigar ruled against the tech giant's motion to dismiss a copyright infringement lawsuit initiated by authors, including Abdi Nazemian. The authors claim that NVIDIA's AI models, such as the NeMo Megatron, were trained using the Books3 dataset, which allegedly contains pirated works from the 'shadow library' Bibliotik. The court's ruling highlights that specific scripts NVIDIA provided to clients for downloading and processing these datasets have no other purpose than to facilitate infringement, a critical point under the new standards established by the recent Supreme Court ruling in Cox v. Sony. This decision marks a pivotal moment for the AI/ML community as it underscores the potential legal repercussions of using questionable datasets for training AI models. NVIDIA's contention that its NeMo framework serves legitimate non-infringing uses was dismissed by the court, which emphasized that the scripts' primary function was to expedite copyright violations. The ongoing lawsuit against NVIDIA, along with similar cases emerging against other tech companies like Meta, indicates a growing scrutiny on the ethical implications of data sourcing in AI development, with repercussions that may shape the industry's practices moving forward.