Generative AI leader Anthropic is once again mired in a copyright dispute. Recently, a group of music publishers led by Concord Music Group and Universal Music Group formally filed a lawsuit with the court, accusing Anthropic of large-scale, systematic "open piracy" during the training of its AI models.

The plaintiffs stated in a statement that Anthropic obtained more than 20,000 copyrighted songs, including sheet music, lyrics, and musical works, through illegal downloads. According to reports, the compensation amount for this lawsuit may exceed $3 billion, which not only sets a new record for copyright litigation in the AI field but will also become one of the largest non-class copyright lawsuits in U.S. history.

The legal team involved in this lawsuit previously represented authors in a copyright case against Anthropic. In an earlier case, "Bartz v. Anthropic," the judge considered that using copyrighted content to train a model may constitute "fair use," but clearly pointed out that if the data was obtained through illegal piracy, it would not be protected by law.

Publishers claim that during the previous discovery phase, they found records of Anthropic using illegal torrents to download thousands of additional works. Although the plaintiffs had attempted to add these allegations to the previous case, the court dismissed them on the grounds of insufficient timely investigation. Therefore, the publishers have chosen to file this high-value lawsuit separately and have also listed Anthropic's CEO and co-founder as defendants.

The language used in the complaint is extremely severe, directly pointing out that Anthropic's corporate image and actual actions are seriously inconsistent. Publishers believe that Anthropic has always presented itself as a "AI safety and research company" with a positive image, emphasizing compliance and ethics, but its multi-billion-dollar business empire is essentially built on the illegal acquisition and piracy of a large amount of cultural heritage.