A federal judge recently determined that Humanity, an AI company, did not violate laws when training its chatbot models using millions of copyrighted books. However, it still faces legal challenges regarding how it obtained these books, specifically from a so-called “Shadow Library” of pirated copies.
In a ruling announced late Monday, San Francisco US District Judge William Alsup stated that it represents a “typical transformation,” as AI systems summarize various works to create original texts that can be deemed “fair use” under US copyright standards.
“Like aspiring writers, Humanity’s AI language model was trained to not simply replicate existing texts,” he noted.
Nonetheless, Alsup dismissed significant claims from a group of authors who had previously sued the company for copyright infringement. He confirmed that Humanity must face trial in December regarding the alleged theft of their works.
“Humanity had no authority to utilize pirated material in the Central Library,” Alsup remarked.
The trio of authors—Andrea Burtz, Charles Graeber, and Kirk Wallace Johnson—argued last summer that Humanity’s actions amounted to “massive theft,” asserting that the company aimed to profit from the human creativity inherent in those works.
As legal proceedings have unfolded in federal courts in San Francisco over the past year, revealed documents indicate internal concerns within Humanity about the legality of using a repository filled with pirated content. This led the company to eventually shift its strategy, opting to purchase copies of digitally available books.
Alsup noted, “The fact that Humanity later bought a copy of a book it initially stole online won’t absolve it of theft but may influence the extent of any statutory damages.”
This ruling could have implications for similar lawsuits involving companies like OpenAI, creators of ChatGPT, and Meta Platforms, the parent company of Facebook and Instagram.
Founded in 2021 by a former OpenAI executive, Humanity is positioned as a developer of a more responsible and secure AI that can draft emails, summarize documents, and facilitate natural conversations.
Nonetheless, a lawsuit filed last year claimed that Humanity’s actions contradicted its stated noble intentions by using a trove of pirated books to develop its AI products.
On Tuesday, Humanity expressed satisfaction with the judge’s acknowledgment that AI training aligns with copyright principles intended to foster creativity and scientific progress. However, they did not address any allegations of copyright infringement.
The authors’ attorney chose not to comment on the matter.
