A district court judge’s partial finding of fair use in a copyright infringement case brought by a group of authors against AI developer Anthropic in June sent shockwaves through the technology and copyright legal worlds. But the court’s subsequent ruling to certify the remaining charge against Anthropic over its downloading millions of pirated books for training as a class action has set off a veritable earthquake.
The designated class could contain as many as seven million putative claimants, and with potential statutory damages of up to $150,000 per work, Anthropic could be on the hook for tens of billions of dollars should the court find in favor of the class plaintiffs.
On July 31, Anthropic filed a petition with the Ninth Circuit Court of Appeals for permission to appeal the district court’s class certification. Now, technology industry groups and consumer and advocacy organizations have separately petitioned to join the case as amici curiae on behalf of Anthropic.
The two groups of amici raise different issues with the district court’s certification ruling, but each frames the case in near-existential terms. “District courts have a duty to take a hard look at class certification,” the technology groups said in their filing. “This case illustrates exactly why. A comparatively straightforward lawsuit filed by three individual authors rapidly has morphed into an unprecedented and gargantuan class action that threatens an entire industry with ruinous financial liability. “
The groups’ brief warned that such potential liability would likely chill investment in the sector. “If there were indeed seven million claimants, statutory damages of $150,000 per work could exceed the entire gross domestic product of each of 171 countries, including Switzerland,” they wrote. “Even a small risk of incurring such extraordinary damages is coercive enough to force almost any defendant to settle. And in this case, given the immense amount of material required to train large language models and the liability risk posed by erroneous class certification, allowing this certification to stand undoubtedly will chill investment and innovation in AI. “
Read more: EU Publishes Mandatory Template for Disclosing AI Training Data
The consumer groups’ petition approaches the problem with the size of the putative class a by emphasizing the lack of common interests and the difficulty of identifying all rights owners who could be entitled to claim a portion of any settlement or judgment in favor of the class.
“As our proposed brief explains, the district court in certifying this class engaged in almost no analysis of the contents of the books at issue in this suit, nor the interests of their copyright holders,” they wrote. “Because class certification turns on whether ‘there are questions of law or fact common to the class,’ understanding the complex and messy rights issues in this case is critical for this court to make a well-informed decision.”
The groups point out that many of them have first-hand experience with class action copyright cases, including the widely referenced Google Books case (Authors Guild v. Google), and suggest it offers a cautionary tale.
“That suit spanned over ten years and involved multiple attempts at class-wide settlement,” they note. “Amici have firsthand experience through that suit with the problems that inadequate representation can cause for absent rightsholders whose views sharply diverge from those who claim to represent them… We seek to help this Court understand why the district court’s failure to adequately consider these deep conflicts within the class is so significant as to constitute manifest error.”
The Ninth Circuit has yet to rule on the various petitions.
