Music publishers sue Anthropic over 'blatant theft' of copyrighted works
Sony and Warner allege the AI firm used lyrics and compositions to train Claude without permission.
Music publishers have filed a lawsuit against AI developer Anthropic, alleging the company engaged in the unauthorized use of copyrighted musical works to train its generative models. The legal action marks a significant escalation in the conflict between intellectual property holders and the developers of large language models (LLMs).
According to the legal filings, the lawsuit accuses Anthropic of "blatant theft" of intellectual property. The litigation is led by major industry players, including Sony and Warner, who claim that their copyrighted lyrics and musical compositions were ingested without permission or compensation to train the Claude AI model. The publishers argue that this unauthorized use of their catalogs constitutes a direct violation of copyright law.
The broader copyright conflict
This litigation is part of a widening trend of legal challenges facing the AI industry. Over the past year, a growing coalition of copyright holders—ranging from novelists and journalists to music publishers—has targeted AI developers. These groups argue that the process of "scraping" the internet for training data is not fair use, but rather a systematic appropriation of creative labor to build commercial products that may eventually compete with the original creators.
Implications for AI training
The outcome of this case could establish a critical legal precedent for the entire generative AI sector. If the courts rule in favor of the publishers, AI companies may be forced to abandon the practice of using unlicensed public data for training. Such a shift would likely mandate a transition toward paid licensing models, where AI firms must negotiate financial agreements with rights holders before incorporating creative content into their models.
What to watch
As the case proceeds, the industry will be watching for how the court defines the boundary between transformative use and copyright infringement in the context of LLM training. It remains to be seen whether Anthropic will seek a settlement with the publishers or challenge the claims in court by arguing that the training process does not create a derivative work of the original music. Further details regarding the specific datasets used by Anthropic are expected to emerge during the discovery phase of the trial.