Judge rejects bid to block Anthropic from using song lyrics for AI training

What just happened? There has been a victory for an AI company in one of the many legal fights over copyrighted works, but the case is far from over. A California judge has rejected a preliminary bid from Universal Music Group (UMG) and other publishers to block their lyrics from being used by Anthropic to train its chatbot, Claude.

UMG, Concord and ABKCO sued Anthropic in 2023 over claims that it used lyrics from at least 500 songs by musicians including Beyonce, the Rolling Stones and the Beach Boys, to train Claude without permission. The lawsuit claimed Anthropic was violating US copyright laws with its actions.

The music groups said they were not asking Anthropic to retrain its existing AI models or remove any from the marketplace. The injunction would not have impacted any models under development, either – it would have just stopped the start-up from using the song lyrics.

Judge Eumi K. Lee said this week that the scope of the requested injunction was too broad as the details remain poorly defined – the injunction would have potentially covered hundreds of thousands of songs, not just the 500 identified in the suit. The judge added that the publishers failed to show that Anthropic’s actions caused them “irreparable harm,” a requirement for the injunction.

However, the court issued two discovery orders granting the publishers significant investigative tools to potentially improve their legal arguments, writes Music Business Worldwide.

The judge also suggested that Anthropic might still have to pay a large amount of damages to the music industry, which the publishers say they are going to pursue “vigorously.”

“Though the Court denied the remainder of our motion, which would have required Anthropic to refrain from training on our lyrics while the case proceeds, it did so on the narrow grounds that our damages for this injury can ultimately be compensated with money damages, which we intend to pursue vigorously,” the publishers said in a statement.

The publishers scored their own victory in January when the court approved protective guardrails for AI-generated song lyrics, preventing Claude from reproducing these lyrics in its outputs to users.

Legal challenges against AI companies for using copyrighted material continue to be fought in the courts. In August last year, two music-generating startups argued that scraping copyrighted tracks from the internet for training purposes was “fair use.”

In August last year, authors sued Anthropic for allegedly using pirated copyrighted work to train Claude.

It’s not just musicians who are aggrieved by what AI companies are doing. Earlier this month, hundreds of stars and Hollywood executives signed an open letter urging the Trump administration to deny proposals from AI companies that would allow their systems to be trained on copyrighted work without obtaining permission.

In February, more than 1,000 musicians, including Kate Bush, Tori Amos, and the Eurythmics’ Annie Lennox released a silent album in protest against proposed changes to Britain’s copyright laws. Mirroring the US, the proposals would allow companies to use artists’ work to train their AI models without permission.