Judge Orders Meta to Defend Against Copyright Infringement Claims in Llama AI Training
A recent court ruling mandates that Meta Platforms, the parent company of Facebook, must answer claims it allegedly removed copyright management information (CMI) from works used to train its Llama family of large language models. The decision, handed down by Judge Vince Chhabria, stems from a case filed in July 2023. The plaintiffs, authors Richard Kadrey, Sarah Silverman, and Christopher Golden, allege that Meta’s use of their copyrighted works for AI training was illegal.
The plaintiffs’ allegations, revealed in January 2025, assert that Meta knew its AI models would generate outputs containing CMI – such as copyright details, licensing terms, and creator information. The plaintiffs contend that Meta subsequently scrubbed this data from the training materials to obscure the copyrighted origins of its AI model’s outputs. This, they argue, violates the US Digital Millennium Copyright Act (DMCA).
In his ruling, Judge Chhabria stated that the plaintiffs’ claims raise a “reasonable, if not particularly strong, inference” that Meta removed CMI to prevent the Llama models from revealing their training data sources. He highlighted that this alleged action constitutes an “identifiable (alleged) infringement.” This ruling increases the likelihood of a settlement or a trial in the case.
Meta has acknowledged using a dataset called Books3 for training its Llama 1 large language model, and the dataset is known to contain copyrighted material.
However, the court did dismiss one claim by the plaintiffs, which alleged that Meta’s use of unlicensed books obtained from peer-to-peer torrents violated California’s Comprehensive Computer Data Access & Fraud Act (CDAFA).
Edward Lee, a law professor at Santa Clara University, commented that Judge Chhabria’s recent decisions are not to be interpreted as definitively determining the issue of fair use. “What it does show is that the plaintiffs’ attorneys were able to find a more particularized factual basis for their DMCA claim, which had been dismissed earlier in the case,” Lee said.
The judge’s ruling, which allows the CMI claim to proceed, adds to the growing legal scrutiny of how copyrighted material is used to train AI models. The Thomson Reuters legal victory last month against Ross Intelligence, an AI firm, set a precedent. The case, which prevents the defendant from claiming fair use to avoid liability, highlighted the legal challenges of using AI-generated outputs, particularly when these outputs resemble an author’s work verbatim, as copyright infringement is more likely to be proven.
Other AI-related copyright litigation is also gaining traction. The case of Tremblay et al vs. OpenAI et al, which was amended last week, seeks to revive previously dismissed DMCA claims by citing new evidence. The amended complaint alleges that OpenAI also removed CMI from its large language models.
Meta did not immediately respond to requests for comment.
