Request to Lift UMG’s Educational Ban on AI Teaching Granted by Judge in Anthropic Trial, Allowing Music Publishers to Gather More Evidence from the Platform
The federal judge in California denied the request from Universal Music Group and other music publishers to block AI company Anthropic from using songs in training its models. This decision marks a significant development in the ongoing copyright battle between rights holders and artificial intelligence developers.
Judge’s Ruling and Concerns Over Licensing
In an order published on Tuesday (March 25), Judge Eum K. Lee of the US District Court in the Northern District of California rejected the publishers’ petition for a preliminary injunction that would ban Anthropic’s “use of copies of texts controlled by publishers for future training of Anthropic AI models.”
The lawsuit, filed in 2023, alleges that Anthropic trained its AI Chatbot Claude using texts from at least 500 songs by artists including Beyoncé, Rolling Stones, and Beach Boys without permission.
In his decision, Judge Lee noted that the publishers failed to demonstrate “irreparable harm,” which is required for such a trial.
“If other artificial intelligence developers receive licenses for the use of material protected by copyright for educational purposes, then from this it follows that … any harm arising from the developing market of the AI licensing will be compensated (through damages), and not irreparable.”
Judge Eum K. Lee, March 25 order
The court expressed concern about the scope of the requested injunction, which would not only cover the 500 works specified in the lawsuit but could potentially extend to “hundreds of thousands” of songs.
Legal Implications and Future Proceedings
The judge emphasized that this decision does not address whether training generative AI models with copyrighted materials constitutes a violation or falls under “fair use,” which remains an unresolved issue for future proceedings.
Importantly, this ruling only addresses one aspect of the publishers’ legal claims against Anthropic and does not affect a previous favorable decision for the publishers.
The case consists of two distinct components:
First, whether copyrights exist in Claude’s outputs (the results seen by users interacting with the AI);
Secondly, whether Anthropic can use copyrighted texts as training data (i.e., creating AI models).
Previously, in January 2025, the publishers achieved a significant victory when Anthropic agreed to implement “guardrails” to prevent Claude from reproducing copyrighted material in its outputs. This agreement remains intact, unchallenged by this new ruling.
In a positive development for UMG and its fellow publishers, the court issued two related orders that could bolster their legal arguments. Within these orders, Magistrate Susan Van Kalen mandated Anthropic to produce a “statistically significant sample” of data from a six-month period covering September 2023 to March 2024.
This wider search approach may provide more evidence of potential violations, enhancing the publishers’ case.
Back in January, UMG expressed satisfaction with the court’s approval of “fences” requiring Anthropic to implement protective measures in its AI models to safeguard against unauthorized reproduction of copyright material. The outcome of the Anthropic case is part of larger legal battles concerning the use of copyright in AI training, with implications for both creative industries and the development of AI technologies.
MBW reached out to Universal Music Group for comments on the latest decision.
Lucas focuses on the interaction between music and society. He specializes in how music influences and reflects cultures, analyzing the role of sound in social, political, and cultural events around the world.
