District Court for the Northern District of California Grants Adobe Motion to Dismiss Nemotron Copyright Claims
Bot Mutiny |
A federal judge dismissed claims that Adobe infringed copyrights by training NVIDIA Nemotron models on pirated books, citing a lack of specific evidence regarding the use of Anna's Archive.
The United States District Court for the Northern District of California has granted a motion to dismiss a portion of a copyright infringement lawsuit against Adobe Inc. The order, filed August 20, 2026, focuses on Adobe's alleged role in developing generative artificial intelligence models using pirated datasets. The case, Tanzer v. Adobe Inc., was first reported by CourtListener. Author E. Molly Tanzer alleged that Adobe unlawfully used her copyrighted books to train two different AI product lines: the SlimLM series and the NVIDIA Nemotron series. Judge Jacqueline Scott Corley dismissed the claims specifically related to the Nemotron models. The court found that Tanzer failed to provide enough facts to support an inference that Adobe acquired her specific works from a shadow library to train those models. ## The Shadow Library Allegations Tanzer's complaint describes shadow libraries as illicit online repositories that defy copyright laws to offer millions of pirated books. The filing identifies several of these entities, including Library Genesis, Z-Library, Sci-hub, and Anna's Archive. According to the complaint, Adobe's SlimLM models were trained on the SlimPajama dataset. This dataset contains the Books3 dataset, which the creator publicly admitted represents the entirety of the shadow library Bibliotik. Books3 includes approximately 196,640 books, including Tanzer's copyrighted titles. The dispute regarding the Nemotron models is more complex. NVIDIA released the Nemotron-4 15B model in February 2024. While NVIDIA didn't list specific datasets, it indicated the model was trained on 8 trillion tokens, with 240.6 billion tokens coming from English-language books. Tanzer estimated this accounts for roughly 1,881,250 books. ## The Nexus Between Adobe and NVIDIA Adobe announced a partnership with NVIDIA in March 2023 to create generative AI models. In March 2024, Adobe stated it was building datasets to train and tune next-generation NVIDIA foundation models, including Nemotron LLMs. Tanzer alleged that Adobe used its PDF Extract tool to build these datasets. The complaint argued that obtaining nearly 1.9 million books for training could only be possible by pirating works from sources like Anna's Archive. The court rejected this logic for the Nemotron-related claims. The order states that Tanzer didn't sufficiently allege Article III standing or a direct infringement claim because she didn't show Adobe actually acquired her works from Anna's Archive for the Nemotron project. ## Distinguishing SlimLM from Nemotron The court's decision creates a distinction between the two AI products mentioned in the suit. The SlimLM claims involve a dataset (Books3) with a documented link to a specific shadow library. The Nemotron claims, however, relied on the sheer scale of the training data as evidence of piracy. The judge ruled that drawing inferences in the plaintiff's favor wasn't enough to sustain the Nemotron portion of the suit. The court found the allegations lacked the specific facts necessary to prove Adobe's training process for Nemotron involved Tanzer's books. Adobe provides tools for image, audio, and video editing and has integrated AI capabilities into its software suite recently. This case is part of a broader legal conflict over how tech companies source the massive amounts of data required to build large language models.