Anthropic agrees to block copyrighted song lyrics from Claude
Music publishers including Universal Music Group, Concord Music Group and ABKCO Music settled a copyright lawsuit against Anthropic over Claude's use of copyrighted song lyrics in training and output. Anthropic agreed to implement safeguards preventing the model from outputting lyrics from copyrighted songs or derivative works, though the parties remained divided over whether Anthropic can use unlicensed lyrics to train future models.", "event_date": "2025-01-06", "facts": [ { "claim": "Universal Music Group, Concord Music Group, ABKCO Music and other record labels jointly sued Anthropic's Claude chatbot for using copyrighted song lyrics without authorization to train the model and outputting those lyrics to users.", "source_url": "https://www.ithome.com.tw/news/166816" }, { "claim": "According to documents filed with the court, Anthropic implemented guardrails in existing AI models and products and committed to maintain similar measures in future models to prevent the model from copying, distributing, displaying or creating derivative works based on copyrighted lyrics.", "source_url": "https://www.ithome.com.tw/news/166816" }, { "claim": "The agreement addresses only model output and does not fully resolve the dispute over training data, with publishers continuing to assert Anthropic should not use unauthorized lyrics to train future models.", "source_url": "https://www.ithome.com.tw/news/166816" }, { "claim": "Anthropic committed to respond quickly and investigate if publishers discover the safeguards fail to prevent copyright infringement, and to work with publishers to resolve issues.", "source_url": "https://www.ithome.com.tw/news/166816" } ] }
What happened
Universal Music Group, Concord Music Group, ABKCO Music and other record labels sued Anthropic, alleging Claude was trained on copyrighted song lyrics without authorization and output those lyrics in response to user queries.
According to court filings reviewed by The Hollywood Reporter, Anthropic agreed to implement guardrails in its existing and future AI models to prevent outputting copyrighted song lyrics or new lyrics based on copyright material.
Reported impact
The record labels claimed the unauthorized use of copyrighted lyrics to train the model and output to users seriously violated their rights.
Index analysis
The agreement addresses only model output and does not fully resolve disputes over training data; the publishers continued to assert Anthropic should not use unauthorized lyrics to train future models, and the parties remained divided on this key issue.
Governance controls
Anthropic committed to respond quickly and investigate if publishers found the safeguards ineffective, and to work with publishers to resolve problems.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.