Artificial intelligence firm Anthropic will pay $1.5 billion to settle a lawsuit filed by authors who accused the company of illegally using their books to train its AI models, a “landmark” agreement that could influence numerous other pending cases over the new technology.
In court filings on Friday (Sept. 5), attorneys for both sides disclosed the terms of an agreement first announced last week. They said the payout, if approved by a federal judge, would be the largest ever from a copyright lawsuit: “This result is nothing short of remarkable.”
The settlement came as major music companies litigate their own infringement suits against AI firms, including one against Anthropic. Such cases, filed by media companies and artists across the creative spectrum, raise make-or-break issues for the booming industry.
AI models like Anthropic’s Claude are “trained” by ingesting millions of earlier works, teaching the machine to spit out new ones. Amid the meteoric rise of the new technology, dozens of lawsuits have been filed in federal court over that process, arguing that AI companies are violating copyrights on a massive scale.
In the current case, filed as a class action by three book authors, a judge ruled this summer that Anthropic merely using existing works to train an AI model was legal under copyright’s “fair use” defense — an important victory for AI companies. But he also found that Claude had been trained with more than seven million pirated books downloaded illegally, and that using those books amounted to infringement.
That ruling, which could have resulted in more than $1 trillion in damages, quickly pushed Anthropic to settle. Under the terms disclosed Friday, Anthorpic will pay $1.5 billion to the authors of more than 500,000 books. It also requires the company to delete the books it acquired illegally and only releases it from past liability.
Related
While the settlement is far less than the maximum that could have been awarded at trial, attorneys for the authors stressed in Friday’s filings that it avoids the risk that Anthropic could have prevailed on its defenses or later overturned the case on appeal.
“Given the very real risk that plaintiffs and the class recover nothing — or a far lower amount — this landmark $1.5 billion+ settlement is a resounding victory for the Class,” the class lawyers wrote.
The lawyers who litigated the case will get a cut of the deal. In Friday’s filings, they did not say how much they would request, but said it could reach as high as 25 percent — meaning $375 million.
One practical impact of the settlement is that the authors will not file an appeal challenging the judge’s ruling this summer that declared AI training a “fair use” — a major victory for Anthropic and other AI companies that will no doubt cite that ruling in their future filings.
“In June, the District Court issued a landmark ruling on AI development and copyright law, finding that Anthropic’s approach to training AI models constitutes fair use,” Aparna Sridhar, Anthropic’s deputy general counsel, said in a statement Friday. “Today’s settlement, if approved, will resolve the plaintiffs’ remaining legacy claims. We remain committed to developing safe AI systems that help people and organizations extend their capabilities, advance scientific discovery, and solve complex problems.”
