Concord Music Group, Inc. and a group of music publishers sued Anthropic PBC in the US District Court for the Northern District of California (Case No. 24-cv-03811-EKL), with the action filed in 2024. The publishers brought copyright claims tied to song lyrics, alleging that Anthropic’s Claude chatbot was trained on, and could reproduce in its outputs, the publishers’ copyrighted musical compositions and lyrics. The case ran in parallel with the recording industry’s separate suits over master recordings at 2024-riaa-v-suno-udio.
The case produced an early, concrete mechanism for handling AI outputs. On January 2, 2025, the court entered a joint stipulation in which Anthropic agreed to maintain its already-implemented “guardrails” - filters intended to stop Claude from reproducing the publishers’ lyrics - and to apply consistent guardrails to new models and products. The stipulation set up a notification process: if the publishers believed the guardrails were failing, they could notify Anthropic in writing, triggering an investigation and a written response. The order resolved that part of the publishers’ preliminary-injunction request while leaving the larger training-data claims for later.
The publishers’ broader bid to block Anthropic’s alleged use of their lyrics in training fared worse. In a decision signed March 25, 2025 (reported at 772 F.Supp.3d 1131), Judge Eumi K. Lee denied the remaining preliminary-injunction motion, holding that the requested relief was too broad and that the publishers had not shown a likelihood of irreparable harm from either reputational or market-related harm. The denial left the underlying copyright claims to proceed; as of the verified date the case remained ongoing.
Why business readers should care: Concord introduced “guardrails” as a litigated, court-supervised remedy - an agreement about how a model behaves at output time rather than a ruling on the legality of training. It showed that AI copyright disputes can be partly managed through technical output controls while the harder training questions, the same ones at issue in 2025-bartz-v-anthropic, remain unresolved.