Sony Music and Warner Music have escalated their legal challenge to artificial intelligence development by launching a copyright infringement lawsuit against Anthropic in California federal court, marking the latest major confrontation between traditional media rights holders and the expanding AI industry. The complaint, lodged on Friday, accuses the AI startup of systematically pirating hundreds of song lyrics and musical compositions from renowned artists including The Beatles, Taylor Swift, and Michael Jackson to fuel the machine learning process behind its Claude conversational model.
The dispute exemplifies a broader pattern of legal hostilities that has emerged as copyright owners—spanning authors, publishers, music labels, and news organisations—challenge technology companies over their use of protected creative works in developing AI systems. This particular case adds significant momentum to an already crowded docket of litigation challenging the legality and ethics of using copyrighted material without explicit permission or compensation. The music industry, which faces distinct challenges from artificial intelligence that can generate synthetic song lyrics and compositions, has become particularly aggressive in defending its intellectual property rights against what it views as systematic exploitation.
Anthropiс's situation is complicated by the fact that Universal Music Group had already initiated legal action against the company in 2023, alleging similar unauthorised use of copyrighted song lyrics in AI training. That case remains unresolved and pending, suggesting that the company may face multiple simultaneous legal battles on the same core issue. The accumulation of lawsuits signals that Anthropic's approach to training data acquisition has drawn sustained scrutiny from multiple sectors of the creative industries.
What makes Anthropic's position particularly vulnerable is its prior settlement with authors in 2024, when the company agreed to pay $1.5 billion to resolve a class action lawsuit over its use of literary works in training datasets. Rather than deterring future infringement claims, that settlement appears to have emboldened rights holders, who now view the payment as merely a business expense that failed to address the fundamental problem. Sony and Warner pointedly argue that Anthropic has factored such settlements into its financial calculations, treating copyright violations as a predictable cost rather than a practice to be discontinued.
The publishers claim that Anthropic obtained their intellectual property through torrent downloads—a deliberate method of acquiring content without authorisation—to train Claude. More provocatively, they allege that the resulting AI model can reproduce copyrighted song lyrics with verbatim accuracy when prompted, demonstrating that the training process created functional copies of protected works within the AI's operational memory. This capability directly threatens the market for the original works and the livelihoods of songwriters and publishers who depend on licensing fees and royalties.
The lawsuit emphasises that Anthropic has weaponised the pirated content to enable Claude to generate what the complaint describes as "vast quantities of purportedly 'new' AI-generated song lyrics." This represents a particularly acute threat to music publishers because it suggests that the AI system can create plausible alternatives to human-composed works, potentially substituting for legitimate copyrighted material in the marketplace. For songwriters and music publishers dependent on licensing revenue, the existence of Claude as a free alternative source of lyrics poses an existential business threat.
The financial stakes in this dispute are substantial. Sony and Warner are seeking damages of up to $150,000 per infringed copyright, which could accumulate to astronomical figures given that they claim hundreds of songs were misappropriated. Beyond monetary compensation, the publishers are requesting injunctive relief—a court order prohibiting Anthropic from continuing to use their works, which would require the company to modify or retrain Claude without the disputed content. Such a remedy could prove technically and financially burdensome to Anthropic.
The valuation disparity that Sony and Warner highlight in their complaint underscores the asymmetry they perceive in the current arrangement. They note that despite allegedly building much of its market value through copyright infringement, Anthropic has achieved a $2 trillion valuation—a figure that dwarfs the $1.5 billion settlement with authors. This argument suggests that the company has profited handsomely from practices that, if judged unlawful, have generated wealth vastly exceeding the penalties imposed. From the music publishers' perspective, this creates a perverse incentive structure in which AI companies profit from theft more than they lose from settlements.
For Southeast Asian readers and observers, this case illuminates the vulnerability of creative industries in developing regions to AI-driven disruption. Many countries across Asia produce substantial original music and literature, yet lack the legal resources and leverage of established Western publishers to defend their rights in American courts. If Anthropic and similar companies can freely appropriate content to build commercially valuable AI systems, creators in smaller markets face disproportionate risk. The outcome of this lawsuit will significantly influence whether developing-world intellectual property can be protected in an era of AI-powered content generation.
The silence from Anthropic, Sony, and Warner on the day the complaint was filed suggests the dispute will be litigated rather than settled quickly. Anthropic may argue that its use constitutes fair use under American copyright law—a legal doctrine that permits some unlicensed use for purposes like criticism, teaching, and research. However, Sony and Warner will counter that commercial training of a profitable AI system is unlikely to qualify as fair use, particularly when the AI generates competing content. The outcome could establish precedent affecting how all AI companies source and utilise training data globally.
