BMG Sues Anthropic, Claiming Unauthorized Use of Its Music Catalog for AI Training

BMG Rights Management’s complaint alleges that Anthropic incorporated copyrighted works from its catalog—including songs by Bruno Mars, the Rolling Stones and other high‑profile artists—into the training data for Claude, the firm’s flagship large‑language model. According to a Reuters filing, the lawsuit claims Anthropic “systematically scraped” the publisher’s digital assets without securing licenses, thereby violating U.S. copyright law and the terms of BMG’s distribution agreements (Reuters, “BMG sues Anthropic for using Bruno Mars, Rolling Stones lyrics in AI training”). The filing seeks an injunction to halt further use of the disputed material, monetary damages, and a court order that Anthropic disclose the scope of the data it harvested. BMG’s legal team argues that the unauthorized ingestion of its music catalog not only deprives songwriters of royalties but also sets a precedent that could erode the economic foundations of the music publishing industry.

Anthropic, which counts Microsoft and Amazon among its investors, has not yet responded publicly to the allegations. In a brief statement to Computerworld, the company said it “takes intellectual‑property concerns seriously” and is reviewing the complaint, but declined to comment on the specifics of its data‑collection practices (Computerworld, “Music giant BMG sues Anthropic over AI training”). Industry observers note that the case arrives at a moment when generative‑AI firms are under increasing scrutiny for how they source training material. A recent antitrust roundup by Reuters highlighted that several AI startups are facing parallel lawsuits from media companies, suggesting a broader regulatory push to enforce clearer licensing frameworks for AI‑generated content.

The dispute underscores a growing tension between AI developers and rights holders over the definition of “fair use” in the context of machine‑learning models. Legal scholars cited by Reuters point out that while U.S. courts have historically allowed limited copying for research, the scale and commercial intent of large‑language models may exceed traditional fair‑use boundaries. If BMG succeeds, the ruling could compel AI firms to obtain explicit licenses for any copyrighted works used in training, potentially reshaping the economics of model development. Conversely, a dismissal could reinforce the industry’s current practice of relying on publicly available data, leaving publishers to seek alternative enforcement mechanisms such as contractual bans or technological safeguards.

For music publishers, the stakes are immediate. BMG represents a catalog that generates roughly $1 billion in annual royalties, according to its own disclosures, and its roster includes legacy acts whose earnings depend on precise accounting of usage. The lawsuit also mentions that Anthropic’s model can reproduce lyrical fragments with a fidelity that could substitute for original recordings in downstream applications, raising concerns about downstream infringement. As Reuters notes, the case could force AI developers to redesign data pipelines, implement stricter provenance tracking, or negotiate blanket licensing deals with rights collectives—steps that would add cost and complexity to an already capital‑intensive sector (Reuters, “Intellectual Property News”).

While the litigation is still in its early stages, the outcome may set a benchmark for how generative‑AI companies engage with copyrighted content across all media types. BMG’s aggressive legal posture signals that music publishers are prepared to defend their assets in court, rather than rely on voluntary compliance. Should the court grant BMG’s request for an injunction, Anthropic could be compelled to purge large swaths of its training corpus, potentially delaying product rollouts and prompting investors to reassess risk exposure. The case therefore serves as a bellwether for the broader AI ecosystem, where the balance between innovation and intellectual‑property protection remains a contested frontier.