Recently, Authentic Brands Group, the owner of Elvis Presley's music catalog, formally filed a lawsuit against AI company Anthropic. The plaintiff alleges that Anthropic, without permission, used 274 Elvis Presley songs to train its AI models, seeking total damages of up to $41 million.
This lawsuit breaks from the usual practice of rights enforcement led by record industry associations or major record labels, becoming one of the first major legal actions brought directly by the estate of a famous artist against an AI developer. The final ruling could very likely set a general legal precedent for how AI companies may legally use existing music catalogs in the future.
Richard Busch, the plaintiff's lead attorney, stated in the complaint that the plaintiff brought this lawsuit to firmly protect its proprietary rights, and said bluntly that the defendant's entire business model is built on copyright infringement on an unprecedented scale. The complaint further claims that Anthropic's AI model, Claude, has been capable since its inception of generating infringing content for almost every song in various music catalogs on the internet.
The allegedly infringed tracks include Elvis classics such as "Can't Help Falling in Love," "All Shook Up," and "A Little Less Conversation." The plaintiff is seeking damages of $150,000 per infringed track, equivalent to about RMB 1.003 million, with total claims of approximately $41 million, or about RMB 274 million.
The complaint also specifically notes that Anthropic only put in place so-called safeguards to limit the output of infringing music content after the matter came to light. However, these safeguards were ineffective, and the company, knowing they were ineffective, still allowed the model to continue producing infringing music. The plaintiff argues that this fully demonstrates that Anthropic knowingly violated the law and acted with ongoing willful intent to infringe others' music copyrights.
At present, the underlying operating logic of AI music platforms generally involves analyzing massive amounts of recorded works to learn and master songs' timbre, structure, and expressive style, and then allowing users to quickly generate AI music through text or voice commands.
Faced with this technological trend, many musicians have repeatedly expressed concerns publicly. On the one hand, they fear that their music works and voices will be used by AI models without authorization and without compensation; on the other hand, they worry deeply that shoddily produced AI-generated music will eventually completely replace excellent original songs created by humans.
In fact, some AI tools have even openly imitated real, living singers on major music streaming platforms. Just last week, streaming giant Spotify released data showing that over the past year, it has removed about 50 million such low-quality AI-generated tracks from its recommendation feeds.