Last month, a federal judge ruled that potentially millions of writers can join a copyright infringement lawsuit brought against the AI startup Anthropic.
The suit, filed by three authors, accused the Claude chatbot maker of using pirated books downloaded from "shadow libraries" such as LibGen to train its large language models. Upping the ante through the roof, US district judge William Alsup said that the trio's suit can represent every single writer of the some seven million books that Anthropic pirated, exposing it to potentially hundreds of billions of dollars in damages.
Now, Anthropic and industry groups are begging an appeals court to reverse the decision, Ars Technica reports, pleading that it would be the largest copyright class action in history — and, tellingly, that it could "financially ruin" the entire AI industry. And that can't be allowed to happen because, as Anthropic humbly argues, AI is simply too important to fail.
"One district court's errors should not be allowed to decide the fate of a transformational GenAI company like Anthropic or so heavily influence the future of the GenAI industry generally," Anthropic said in a petition filed last week.
The petition calls the decision erroneous and accuses the court of rushing to certify the suit "without any reliable means to identify class members or adjudicate their claims." Alsup, it argues, failed to conduct a "rigorous analysis" and made his decision based "only" on his fifty years of experience, per Ars.
Most of all, Anthropic says that the stakes are so extreme that it would be forced to settle right away, denying it a fair chance to defend how it trained its AI models in court.
"Given the size of the class and risk of statutory damages up to $150,000 per work, Anthropic faces hundreds of billions of dollars in potential liability — a tremendous amount, particularly for an emerging company," it says. (Anthropic regularly receives billions of dollars in investment and was recently valued at $61.5 billion, which is larger than the yearly revenue of the entire US publishing industry.)
The Anthropic case is one of several high-profile suits that could have existential implications for the industry's future. AI companies have depended on gathering training data at little to no cost, and authors and artists have sued the likes of OpenAI and Meta for using their work without permission or compensation. The industry argues that their training counts as "fair use," and have generally struck a similar, pleading tone in defending the practice: that if it was forced to pay up for all the copyrighted art it used, it effectively would sink the entire endeavor.
Such an admission might warrant some reflection on whether hinging an industry's existence on pillaging virtually the entire internet is ethical or sustainable, but instead, tech companies have deployed arguments such as the books they used having no economic value, or invoking the bogeyman of China and claiming that anything that slows down American AI companies' progress poses a national security risk.
An interesting curveball in this case is Anthropic citing precedent that "copyright claims are poor candidates for class-action treatment" because they require that every claimant proves their ownership of a work, which is rarely straightforward in the world of publishing.
... continue reading