American artificial intelligence (AI) company Anthropic, which develops large language models competing with platforms like OpenAI’s ChatGPT and Google’s Gemini, has won a key ruling in a United States federal court. A federal judge ruled this week that AI developers can train AI models on copyrighted content without obtaining permission from the content creators.
As The Verge reports, U.S. Federal Judge William Alsup of the Northern District of California ruled that Anthropic has the legal right to train AI models using copyrighted work. Judge Alsup says that this use falls under fair use.
In the lawsuit, Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson v. Anthropic PBC, the three plaintiffs, all authors, allege that Anthropic had no right to use their protected works to train its family of Claude AI models.
Judge Alsup disagrees, ruling that Anthropic’s use of the plaintiffs’ works, which included buying physical books, stripping them bare, and scanning the text into its training workflow, falls under fair use.
Fair use has long been a crucial defense for AI companies, which require huge libraries of human-created work to sufficiently train their various AI models. Understandably, artists have resisted, and copyright infringement lawsuits have popped up left and right.
Alsup’s ruling is multi-faceted, however. While the federal judge has sided with Anthropic on the matter of using legally acquired, copyrighted materials to train AI models, the judge takes significant issue with some of Anthropic’s other behavior, including storing more than seven million pirated books in a central library. This is not protected under the fair use doctrine, and the judge has set a second trial later this year to determine the damages Anthropic may owe for this infringement. As Reuters reports, “U.S. copyright law says that willful copyright infringement can justify statutory damages of up to $150,000 per work.”
However, potentially even more influential than Judge Alsup’s ruling that training AI on copyrighted material can be protected under the doctrine of fair use is his additional decision that building AI models using copyrighted work can be considered sufficiently transformative to avoid violating copyright.
“To summarize the analysis that now follows, the use of the books at issue to train Claude and its precursors was exceedingly transformative and was a fair use under Section 107 of the Copyright Act,” Judge Alsup writes.
“Anthropic’s LLMs have not reproduced to the public a given work’s creative elements, nor even one author’s identifiable expressive style (assuming arguendo that these are even copyrightable). Yes, Claude has outputted grammar, composition, and style that the underlying LLM distilled from thousands of works,” Alsup continues elsewhere in his ruling. “But if someone were to read all the modern-day classics because of their exceptional expression, memorize them, and then emulate a blend of their best writing, would that violate the Copyright Act? Of course not.”
Alsup calls using legally-acquired copyrighted works to train LLMs as “quintessentially transformative,” claiming that Anthropic is using existing works “not to race ahead and replicate or supplant” the creators, but to “turn a hard corner and create something different.”
In their lawsuit, the plaintiffs alleged that, in general, training LLMs would “result in an explosion of works competing with their works,” as Alsup characterizes it. The judge strongly disagrees with this complaint.
“But Authors’ complaint is no different than it would be if they complained that training schoolchildren to write well would result in an explosion of competing works,” Alsup writes. “This is not the kind of competitive or creative displacement that concerns the Copyright Act. The Act seeks to advance original works of authorship, not to protect authors against competition.”
As a result of Alsup’s ruling, AI companies now have a proven avenue through which they can defend their training work on the grounds of fair use. The ruling also asserts that some training applications are sufficiently transformative to be legally protected. There is little doubt that this new ruling could prove to be a landmark case that influences how other judges handle copyright claims levied against AI companies.
That said, Anthropic will still need to answer for its alleged piracy.
Image credits: Header photo licensed via Depositphotos.
Source link