Music publisher BMG has filed a lawsuit against AI company Anthropic, alleging that the company’s popular chatbot, Claude, was trained on unlicensed song lyrics and now generates infringing content, marking a significant escalation in the ongoing legal battles over artificial intelligence and intellectual property.
BMG Joins Growing AI Copyright Litigation
BMG, a major music company, officially entered the escalating AI copyright dispute by filing a lawsuit against Anthropic, the creator of the AI chatbot Claude. The complaint, lodged by various BMG publishing entities on Tuesday, asserts that Anthropic illegally utilized copyrighted song lyrics to train its AI models. This action places BMG alongside other prominent music rights holders who have initiated legal proceedings against AI developers for alleged copyright infringement.
The core of BMG’s accusation is that Anthropic’s Claude chatbot, when prompted, generates lyrics that are infringing upon BMG’s intellectual property. This includes lyrics from well-known songs that have achieved chart success, such as Ariana Grande’s "7 Rings" and Bruno Mars’ "Uptown Funk."
Allegations of Unlicensed Training and Infringing Outputs
According to the lawsuit, filed by the prominent entertainment law firm Manatt Phelps & Phillips, Anthropic has committed "egregious law-breaking." The legal document states, "Anthropic has blatantly violated the copyright laws and caused direct harm to BMG and the songwriters it proudly represents." The complaint further asserts that while many innovators have developed groundbreaking technologies while respecting copyright, Anthropic’s rapid advancement is not an excuse for its alleged illegal conduct.
BMG’s legal challenge mirrors existing copyright litigation filed against Anthropic by Universal Music Publishing Group (UMPG), Concord Music, and ABKCO Music. These publishers have been engaged in legal battles with Anthropic since 2023. The specific reasons for BMG’s decision to pursue its own independent case, rather than joining the existing litigation, remain unclear. Notably, recent reports indicate that BMG and Concord are reportedly in discussions regarding a potential merger.
A History of Cease-and-Desist and Unresponsiveness
BMG claims to have sent a cease-and-desist letter to Anthropic in December, but the AI company reportedly failed to respond. A spokesperson for BMG emphasized the company’s commitment to protecting the rights of its artists and songwriters. "Building an industry on the backs of our songwriters, recording artists, and producers, without permission or compensation, is never acceptable," the spokesperson stated.
The spokesperson also expressed a belief that generative AI, with proper permissions, could be a valuable tool for enhancing creativity rather than replacing it. However, they underscored that "copyright protection and fair remuneration are non-negotiable." Anthropic has not yet issued a statement in response to the lawsuit.
Dual Infringement Claims: Training Data and Generated Content
Similar to the ongoing lawsuits brought by other music publishers, BMG’s complaint alleges that Anthropic has infringed on its copyrights in two distinct ways: through the data used to train Claude and through the content the chatbot generates. This dual claim distinguishes the publishers’ cases from separate lawsuits filed by major record labels against AI music generators like Suno and Udio, which primarily focus on the issue of unlicensed training data.
Input Side: Allegations of Piracy and Scraped Data
BMG contends that Anthropic has infringed its intellectual property by incorporating copyrighted lyrics into Claude’s "enormous" training dataset. The lawsuit asserts that this training material includes text scraped from BMG-licensed lyric databases, such as MusicMatch and LyricFind, as well as from sheet music collections featuring the works of renowned artists like The Rolling Stones and Justin Bieber.
A significant aspect of BMG’s claim is the allegation that Anthropic acquired much of this training data from illegal pirate libraries through torrenting. This point is particularly salient given a recent judicial ruling in a separate case. A judge determined that Anthropic could be held liable for storing torrented books, a decision that led to a $1.5 billion settlement with authors. This ruling has emboldened other copyright plaintiffs, including music publishers and record labels, to incorporate piracy claims into their existing lawsuits against AI companies.
Output Side: Generative AI Producing Infringing Lyrics
Beyond the training data, BMG also alleges that Anthropic is infringing copyrights through the actual text generated by Claude. The lawsuit claims that when prompted by users, Claude has produced complete or substantial portions of lyrics from multiple BMG-owned compositions. Examples cited include "Uptown Funk," Louis Armstrong’s "What a Wonderful World," and 3 Doors Down’s "Kryptonite."
The complaint further details how Claude generates outputs that include unauthorized copies and derivative works of BMG’s copyrighted lyrics, even when users request "new" or "original" song lyrics. The lawsuit suggests that Claude combines various songs into a single output, effectively creating a mash-up in response to user prompts.
Seeking Substantial Damages and the Future of AI Fair Use
BMG is seeking significant financial damages from Anthropic, arguing that the company has amassed "a fortune built on stolen copyrighted works." The lawsuit highlights Anthropic’s recent $30 billion funding round, which valued the company at $380 billion. BMG is requesting the statutory maximum of $150,000 per act of infringement. With a non-exhaustive list of 467 allegedly infringed songs attached to the complaint, the total damages sought could amount to at least $70 million.
Anthropic, like other AI companies facing a wave of copyright litigation, has maintained that its activities are protected under the legal principle of "fair use." This tenet of copyright law permits the unlicensed use of copyrighted material for transformative purposes. However, the question of whether AI training constitutes fair use remains an unresolved legal issue, currently being litigated across numerous courtrooms nationwide. The outcome of these cases is expected to shape the future landscape of AI development and its interaction with creative industries.
Broader Industry Implications and the AI Copyright Frontier
The lawsuit filed by BMG signifies a critical juncture in the evolving relationship between artificial intelligence and intellectual property rights. As AI technologies become increasingly sophisticated, the legal frameworks governing their development and deployment are being tested. The music industry, in particular, has been at the forefront of these challenges, as its vast catalog of creative works represents a rich source of data for AI training.
The legal battles initiated by BMG, UMPG, Concord, and ABKCO Music underscore a unified front among music publishers in demanding accountability from AI developers. These actions highlight the potential for AI to disrupt traditional revenue streams and creative processes within the music ecosystem. The plaintiffs are not only seeking financial compensation but also aiming to establish legal precedents that will govern the ethical and legal use of copyrighted material in AI development.
The Fair Use Debate in AI
The central argument for AI companies like Anthropic often revolves around the doctrine of fair use. This legal defense allows for the use of copyrighted material without permission for purposes such as criticism, comment, news reporting, teaching, scholarship, or research. AI developers contend that the process of training AI models on vast datasets, including copyrighted works, is transformative and essential for the advancement of AI technology. They argue that the AI does not reproduce the original works in a manner that directly competes with them, but rather uses them to learn patterns and generate new content.
However, copyright holders argue that the unauthorized use of their works for training AI models constitutes a form of mass infringement that undermines the value of their intellectual property. They contend that AI-generated content that closely mimics or incorporates elements of existing copyrighted works can directly harm the market for the original creations. The legal definition of "transformative use" is subjective and will likely be a key point of contention in these ongoing cases.
Impact on Songwriters and Artists
For songwriters and artists, the implications of these lawsuits are profound. Their creative output, often the result of years of dedication and effort, is the raw material that fuels AI models. The lack of compensation or recognition for the use of their work in training AI can be seen as a direct threat to their livelihoods and the sustainability of creative careers. The BMG lawsuit, with its emphasis on the "life’s work" of songwriters, underscores this concern.
The legal victories or defeats in these high-profile cases could significantly influence how AI companies approach data acquisition and development. A ruling in favor of the copyright holders could lead to more stringent licensing requirements and greater financial obligations for AI developers. Conversely, a ruling in favor of AI companies could accelerate AI development but potentially at the expense of creators’ rights.
The Broader AI Landscape
Beyond the music industry, the principles at play in these lawsuits have far-reaching implications for other creative fields, including literature, visual arts, and journalism. As AI continues to evolve, similar legal challenges are likely to emerge across various sectors. The outcomes of these AI copyright disputes will contribute to the ongoing societal conversation about the ethical boundaries of technological innovation and the protection of intellectual property in the digital age. The battle between BMG and Anthropic is not just a legal dispute; it is a crucial chapter in defining the future of creativity and copyright in the era of artificial intelligence.
