AI industry fears largest ever class action copyright lawsuit in the US

AI industry associations are calling on the US Court of Appeals for Northern California to block a class action lawsuit for copyright infringement against the AI company Anthropic. They complain that this legal action, which could potentially be joined by up to seven million plaintiffs, could financially ruin the entire US artificial intelligence industry and jeopardize US technological competitiveness.

The original lawsuit was filed last year by three book authors Andrea Bartz, Charles Graeber and Kirk Wallace Johnson. They claim that Anthropic used their copyrighted works to train its AI models like Claude. After a federal district court allowed the class-action lawsuit to proceed, at least in part, in June, Anthropic recently filed an appeal. The company argues that the lead judge, William Alsup, who has been practicing law in Silicon Valley for around 50 years, did not conduct a thorough analysis when making his decision and underestimated the potential risks of the case.

If the class action is approved, Anthropic fears claims amounting to hundreds of billions of US dollars in damages within a few months. Considering this existential risk, the company could be forced to agree to a settlement, it argues. In principle, the company is entitled to deny the allegations. The Claude operator warns in a that a costly way out-of-court settlement would set a worrying precedent for the entire generative AI industry.

The challenge of rights clearance

IT associations such as the Consumer Technology Association (CTA) and the Computer and Communications Industry Association (CCIA) supported Anthropic in a statement submitted to the court on Thursday. They also expressed alarm that "the district court's flawed class action" would cause immense harm "not just to a single AI company". In general, copyright disputes are considered unsuitable for class actions because each individual author must prove ownership of his or her works.

Even author advocacy and civil rights organizations such as the Authors Alliance, the American Library Association, the Association of Research Libraries, the Electronic Frontier Foundation (EFF) and Public Knowledge support Anthropic's appeal. They point out that the Google Books case has shown that proving ownership of usage and exploitation rights is anything but easy.

The advocates of the creative industries criticize the fact that Judge Alsup considered the identification of up to seven million rights holders to be too simple. In reality, it is extremely complicated to clarify the ownership of so many works, as publishers may no longer exist, rights may have been inherited by individuals or they may be "orphan works" without clear rights holders.

Wrong approach class action

According to the corresponding submission, there is "no realistic way" to solve these problems collectively. An inadequate notification system for potential claimants and the immense complexity of rights clearance would make the process unmanageable. The organizations warn that continuing the lawsuit would only increase uncertainty about AI training with copyrighted material. In addition, some authors might never find out that the lawsuit is even being pursued. This would prevent them from claiming the damages to which they are entitled. It would therefore be better to stop the proceedings.

According to the first instance ruling, the use of book copies for AI training is permitted, but downloading electronic books from "pirate sites" is illegal. Before Anthropic began scanning legally purchased print editions, the company downloaded beyond seven million e-books from the Internet without bothering with licenses. Dozens of other lawsuits are pending in the USA alleging copyright infringements by AI operators.

(nen)