Vancouver-based author J.B. MacKinnon has filed four proposed class action lawsuits against major technology companies, alleging they illegally used copyrighted works by Canadian writers to train artificial intelligence systems, according to the Canadian Broadcasting Corporation.

MacKinnon, author of The 100-Mile Diet and The Once and Future World, serves as the representative plaintiff in separate suits he filed in B.C. Supreme Court against Nvidia, Meta, Anthropic, and Databricks Inc. The cases target what MacKinnon described as unauthorized use of copyrighted material in developing large language models.

The lawsuit claims MacKinnon's books were part of a 196,640-book dataset that Nvidia used without securing licensing fees or author consent. "This isn't a situation where some copyrighted material appears as a small fraction of the larger process," MacKinnon told the CBC. "The models were entirely built on the mining of copyrighted work."

The proposed class actions describe plaintiffs as holders of Canadian copyright in works used by the companies to build their LLMs. All four cases require court certification to proceed.

The lawsuits seek multiple forms of relief, including an injunction to prevent Nvidia from continuing to infringe on the copyrights of Canadian authors. This would stop Nvidia from using or benefiting from the unauthorized use of copyrighted works to train AI models.

Plaintiffs also request an accounting of profits, requiring Nvidia to reveal and potentially disgorge profits attributable to the alleged copyright infringement. The suits seek restitution to copyright owners for compensation regarding the unlawful use of their works.

Additionally, the cases pursue punitive damages aimed at penalizing Nvidia for what the lawsuit calls "repeated, willful and knowing infringement" of copyright.

Beyond the alleged unauthorized use, the lawsuits claim the companies took additional steps to conceal copyright infringement, going so far as to remove copyright information from books before feeding them into the models and then instructing their LLMs to respond misleadingly when asked about the use of copyrighted material.

Similar legal challenges in the U.S. have been met with mixed results. Last month, two judges working out of the same San Francisco courthouse in cases filed against Meta and Anthropic that the while the use of books to train LLM’s are likely to have constituted fair use, the use of pirated material was itself illegal. However, a judge also ruled last month a class action lawsuit filed by three writers against Anthropic can proceed.

Nevertheless, the Canadian courts are not bound to U.S. legal precedent.

Reidar Mogerman, an attorney representing MacKinnon, told CBC that the issue extends well beyond individual cases. "The goal of the companies is not to transform the world, it's to make money," Mogerman said. "I think you can respect both the values of the copyright system and the ability of these companies to create these models.... It's just that you can't throw out one to create the other, especially when the thing you create is going to be a competitive threat to the work the authors did."

MacKinnon, who also serves as an adjunct professor of journalism at the University of British Columbia, characterized the legal effort as challenging established industry practices. "The most disturbing aspect of it is that those large language models and AIs can then compete with human writers," he said, "and are likely to displace human writers.” He added: "Collectively as Canadian writers, we're certainly the David against the Goliath in this case."