AI copyright news, graded by source

Judge Lee denies music publishers a preliminary injunction against Anthropic training on their lyrics, finding no proof of irreparable harm, after the sides agreed to a court-ordered guardrails deal

Judge Lee denied the publishers' renewed motion for a preliminary injunction on March 25, 2025, without reaching the merits of their infringement claims. The publishers, eight companies led by Concord and Universal Music Publishing, asked the court to bar Anthropic from using their lyrics to train future Claude models, citing reputational harm from mashups and other outputs and harm to a licensing market for AI training. The court held the harm evidence was general and conclusory, that the output examples predated the original 2023 motion and were addressed by the parties' stipulation, and that it would not grant extraordinary relief to protect licensing rights not yet established. It distinguished the Delaware ROSS Intelligence ruling. The case was filed in Tennessee in October 2023 and transferred to California in June 2024. On Jan. 2, 2025 the court entered a stipulation under which Anthropic will keep its existing guardrails against lyric output in current and future models, investigate written notices from the publishers that the guardrails are failing, and submit to the court's enforcement, which mooted that part of the motion. The denial was without prejudice.